Prevalence of Sexual Dysfunctions
Results from a Decade of Research
Jeffrey Simons, Ph.D. and Michael P. Carey, Ph.D.*
Arch Sex Behav. Author manuscript; available in PMC 2008 Jun 12.
Published in final edited form as:
Arch Sex Behav. 2001 Apr; 30(2): 177–219.
See other articles in PMC that cite the published article.
Ten years of research that has provided data regarding the prevalence of sexual dysfunctions is reviewed. A thorough review of the literature identified 52 studies that have been published in the 10 years since an earlier review by Spector and Carey (1990). Community samples indicate a current prevalence of 0 - 3% for male orgasmic disorder, 0 - 5% for erectile disorder, and 0 - 3% for male hypoactive sexual desire disorder. Pooling current and 1-year figures provides community prevalence estimates of 7 - 10% for female orgasmic disorder and 4 - 5% for premature ejaculation. Stable community estimates of the current prevalence for the other sexual dysfunctions remain unavailable. Prevalence estimates obtained from primary care and sexuality clinic samples are characteristically higher. Although a relatively large number of studies have been conducted since Spector and Carey’s (1990) review, the lack of methodological rigor of many studies limits the confidence that can be placed in these findings.
Keywords: sexual dysfunction, prevalence, epidemiology, sexuality
Prevalence of Sexual Dysfunctions: Results from a Decade of Research
Sexual dysfunctions are believed to be among the more prevalent psychological disorders in the general population (Spector & Carey, 1990). The sales data and media attention associated with recent biomedical treatments (e.g., Viagra ®) corroborates the commonness of such dysfunctions. Despite their apparent prevalence, however, sexual disorders have typically not been included in large scale epidemiologic studies such as the Epidemiologic Catchment Area (ECA) Study (Regier et al., 1990). Lacking such large-scale epidemiologic data, sexual health practitioners and scientists must attempt to integrate smaller studies to obtain population estimates regarding the prevalence of sexual disorders. To this end, the current paper provides an updated review and methodological critique of the published empirical studies that provide epidemiologic data as a resource for professionals who are investigating the etiology, assessment, or treatment of these disorders as well as for those involved in the allocation of resources for prevention, treatment, and research.
Research on the sexual dysfunctions has increased dramatically since Spector and Carey’s (1990) review. During the last 10 years, there have been 52 empirical studies providing epidemiologic data on sexual dysfunctions (see Table 1). This compares to 47 studies that were published in the 50 years (1940 - 1989) previously. In the current paper, we review these 51 studies published since the Spector and Carey (1990) review. To identify this research, we searched Medline and PsychInfo databases for articles appearing between 1990 and 1999 using the following key words: (epidemiology OR incidence OR prevalence) AND (premature ejaculation OR impotence OR erectile dysfunction OR erectile disorder OR hypoactive sexual desire disorder OR sexual aversion disorder OR sexual arousal disorder OR orgasmic disorder OR inhibited female orgasm OR inhibited male orgasm OR dyspareunia OR vaginismus OR frigidity OR inhibited sexual desire OR anorgasmia OR ejaculatory dysfunction OR erectile incompetence OR aspermatism OR retarded ejaculation OR ejaculatory inhibition OR absence of ejaculation). Studies were excluded if they examined dysfunctions exclusively among surgical patients (e.g., radical prostectomy), examined erectile dysfunction among diabetics (for a recent review see Weinhardt and Carey, 1998), examined sexual dysfunction among patients with cancers (e.g., prostrate or ovarian cancer) or undergoing chemotherapy, or dysfunctions resulting from medication side effects (e.g., selective serotonin reuptake inhibitors). In addition to the database search, the reference sections of the included papers were reviewed for additional studies.
Studies of the prevalence of sexual dysfunction (1990 - 1999)
We review the studies separately by gender, type of sample (i.e., clinical, community, comorbid disorders), and the phase of sexual response cycle (i.e., desire, arousal, orgasm) that was investigated. In reviewing studies of each dysfunction, we first review prevalence in clinical and community samples without comorbid disorders. We then review prevalence rates in subpopulations when these were available.
We have chosen to use the Diagnostic and Statistical Manual of Mental Disorders - IV (DSM - IV; American Psychiatric Association [APA], 1994) nomenclature because it provides a standard set of definitions with which to structure the review. However, only eight of the 51 studies (16%) used the diagnostic criteria found in the DSM. More than a third of the studies did not provide any operational definition of the dysfunction being investigated. Thus, there exists the potential for great variation in the object of investigation within each of the sections that we have labeled according to the DSM-IV nomenclature. We discuss the potential influence of such variation when non-standard or non-existent criteria coincide with extreme prevalence estimates. However, at the outset, it should be noted that lack of standard criteria is an important factor affecting comparisons among the majority of studies reviewed. Terminology utilized by the respective investigators is retained in Table 1. Prior to each section we provide a brief description of the respective dysfunctions as defined in the DSM - IV. It is understood that in addition to the sexual dysfunction the DSM system uses two additional criteria to determine a sexual disorder, (1) the disturbance causes marked distress or interpersonal difficulty, and (2) the dysfunction is not better accounted for by another Axis I disorder (except another sexual dysfunction) and is not due to the physiological effects of a substance or general medical condition. Thus, despite our adoption of the DSM nomenclature the majority of the prevalence data pertain to sexual dysfunctions rather than DSM disorders.
Female Orgasmic Disorder is characterized by unusual difficulty in attaining orgasm. Determination of dysfunction depends upon age, sexual experience, and the adequacy of sexual stimulation. Prevalence rates of female orgasmic disorder range from 4% - 7% across three large population samples (Ernst et al. 1993; Lindal & Staffansson, 1993; Ventegodt, 1998). Each of these studies was conducted in a northern European country. Ernst et al. (1993) and Fugl-Meyer and Sjogren Fugl-Meyer (1999) report 1-year prevalence statistics, Lindal and Stefansson (1993) lifetime prevalence, and Ventegodt (1998) current prevalence. Despite the more restrictive time frame, Ernst et al. (1993), Fugl-Meyer and Sjogren Fugl-Meyer (1999), and Ventegodt (1998) report higher prevalence rates (7 - 10%) than do Lindal and Stefansson (4% lifetime, 1993). An fifth study in the U.S. reports that 24% of women have been unable to reach orgasm for several months over the past year (Laumann et al., 1999). Only Lindal and Stefansson (1993) used DSM criteria whereas the other investigators used idiosyncratic definitions. These studies demonstrate the potential influence of nonstandard diagnostic criteria on prevalence estimates. It is uncertain whether the high estimate obtained by Laumann and colleagues (1999) reflects differences in assessment technique or population differences.
Results from three studies in four primary care populations (Chandraiah et al., 1991; Read et al., 1997; Shahar et al., 1991) estimate the prevalence of female orgasmic disorder between 5% (lifetime) (Chandraiah et al., 1991) and 42% (current) (Read et al. 1997). Chandraiah et al. (1991) was the only study to use DSM-III criteria and reported a 5% lifetime prevalence in 43 women attending a premenstrual syndrome (PMS) clinic. As expected, the prevalence of female orgasmic disorder appears to be higher in primary care settings compared to the general population. However, methodological problems, especially poorly defined criteria, preclude definitive conclusions in this regard. Two studies in primary care settings report high prevalence rates but do not report the criteria used for assessing the dysfunction. In contrast, Chandraiah et al. (1991) report a prevalence estimate based upon DSM - III criteria that is within the range of recent northern European population based estimates.
Prevalence estimates of female orgasmic disorder obtained in sexuality clinics are 0% (Bhui et al., 1994), 22% (Goldmeier et al., 1997), and 41% (Bhui et al., 1994; Jindal & Dhall, 1990). However, Bhui et al. (1994) has a sample of three women and thus can not be taken as a useful estimate of prevalence. Neither Jindal and Dhall (1990) nor Goldmeier and colleagues (1997) used DSM criteria, making these estimates difficult to reconcile.
Female Sexual Arousal Disorder is characterized by an insufficient lubrication-swelling response to sexual excitement. Little research has been conducted on female sexual arousal disorder. Lindal and Stefansson (1993) reported a lifetime prevalence of 6% in a large random population sample. This study used DSM-III criteria enhancing confidence in this estimate. Fugl-Meyer and Sjogren Fugl-Meyer (1999) report a 1-year prevalence of 8% in a large Swedish sample. In contrast, Laumann and colleagues report a 1-year prevalence of 19% in a representative U.S. sample. Lack of standard criteria across studies makes this large disparity difficult to interpret. Chandraiah et al. (1991) reported a lifetime prevalence of 21% in a primary care setting based upon DSM-III criteria. One study examined women attending a sex therapy clinic (Verma et al., 1998). This was a small sample from a North Indian clinic, in which no cases of female dysfunctions were reported. Thus, more research on the prevalence of arousal disorders in women is needed.
Hypoactive Sexual Desire Disorder (HSDD) is defined as deficient sexual fantasies and desire for sexual activity. Determination of dysfunction is relative to age and the context of the person’s life. Prevalence estimates for HSDD among females range from 5% (Ventegodt, 1998) to 46% (Chiechi et al., 1997) across seven studies. The two highest estimates (46%; Chiechi et al., 1997; 37%, Wasti et al., 1993) are from samples of postmenopausal women. The other studies report 1 - year estimates ranging from 14% (Fugl-Meyer & Sjogren Fugl-Meyer) to 33% (Laumann et al., 1999). Only Lindal and Stefansson (1993) used DSM - III criteria, and they reported a significant higher lifetime prevalence among women (16%) than men (4%). The many methodological differences across few studies makes interpretation of this large range difficult.
Six studies examined HSDD among women in primary care settings. Brown and colleagues report a current prevalence of 20% and 31% among HIV+ women (Brown & Rundell, 1993; Brown & Rundell, 1990). It should be noted that these two studies are part of a five year longitudinal study and may reflect overlap of participants between studies. Jamieson and Steege (1996) report a current prevalence of 10% among women in a gynecology clinic. Goggin et al. (1998) reported a positive relation between HSDD and depressive symptoms, low life satisfaction, and perceived risk for HIV infection. This study reported a current estimate of 39% in a sample obtained from the community and HIV/AIDS health clinics. Chandraiah et al. (1991) studied women in a PMS clinic and reported a lifetime prevalence of 21% based upon DSM-III criteria. Chiechi et al. (1997) report a current prevalence of 46% among post-menopausal women. Thus, current estimates range from 10% (Jamieson & Steege, 1996) to 46%.
Dyspareunia is characterized by persistent genital pain during sexual intercourse. Dyspareunia is not diagnosed if the pain is exclusively due to vaginismus or lack of lubrication. Sexual pain disorders have been the focus of a relatively large number of studies. Prevalence estimates range from 3% (Lindal & Stefansson, 1993; Ventegodt, 1998) to 18% (Moody, 1993) in the general population. This relatively large range is difficult to explain in terms of methodological differences. For example, percentages at the high end were reported by Moody (1993) and Glatt et al. (1990), who report point prevalence as opposed to the 3% lifetime prevalence reported by Lindal and Stefansson (1993). Similarly, both high (Glatt et al., 1990; Laumann et al., 1999; Moody, 1993) and low (Ernst et al., 1993; Rekers et al., 1992; Ventegodt, 1998) estimates have been reported based on operational as opposed to DSM criteria. Only Lindal and Stefansson (1993) used DSM-III criteria. The only consistent difference is that the lower estimates are from Northern European countries whereas the higher ones are from the U.S.
In general practice settings, current estimates range from 3% (Heisterberg, 1993) to 46% (Jamieson & Steege, 1996) across six studies. Jamieson and Steege (1996) report point prevalence in a large primary care sample of “pain during or after intercourse”. The high prevalence in this and the Weber et al. (1995) sample (41%) are three times greater than the next highest estimate of 14% (Heisterberg, 1993). These high estimates may be due to the operational definition of the disorder. The DSM - IV diagnostic criteria exclude pain that is associated exclusively with lack of lubrication or vaginismus. Neither of these high estimates exclude pain secondary to these causes. In fact, Weber and colleagues (1995) specifically include vaginal dryness as a sufficient criterion. Chandraiah et al. (1991), using DSM - III criteria, report a lifetime prevalence of 12%. The range of estimates obtained illustrate the differences that result from nonstandard definitions and the importance of clearly specifying the definitions used.
Two studies report prevalence of dyspareunia in sexuality clinic settings. No female dysfunctions were reported in a large sample of women attending a North Indian clinic (Verma et al., 1998). Jindal and Dhall (1990) report a current prevalence of 13% in an infertility clinic. In relation to other sexual dysfunctions, dyspareunia appears to be a less common presenting problem at sexuality clinics. This may be because pain disorders are more likely to present at primary care rather than sexuality specific clinics.
Older women have been a population in which dyspareunia has been a focus of research. There is evidence that dyspareunia is more prevalent in post-menopausal women (Rekers et al., 1992). Prevalence estimates in community samples of postmenopausal women range from 2% (Barlow et al., 1997) to 21% (Wasti et al., 1993). Barlow et al. (1997) report 2-year prevalence data of “painful intercourse”. However, only 19% of the sample reported having penetrative intercourse. Thus, the prevalence of pain during intercourse may be more appropriately estimated at nine percent. Thus, prevalence estimates range from 9% - 21% across five studies (Barlow, et al., 1997; Diokno et al., 1990; Ramoso-Jalbuena, 1994; Rekers et al., 1992; Wasti et al., 1993).
Vaginismus is defined as persistent involuntary spasm of the vagina interfering with intercourse. Prevalence data for vaginismus are scant without the benefit of multiple studies within populations. Read et al. (1997) report a current estimate of 30% in a primary care setting. Community estimates range from 0.5% - 1% (Fugl-Meyer & Sjogren Fugl-Meyer, 1999; Ventegodt, 1998). Verma et al. (1998) report a prevalence of 0% in a sexuality clinic sample. (Goldmeier et al., 1997) reports a current prevalence of 25% in a STD clinic. Thus, no clear estimate emerges.
Male Orgasmic Disorder is characterized by persistent difficulty in attaining orgasm. Determination of dysfunction depends upon age and the adequacy of sexual stimulation. The limited data available regarding male orgasmic disorder suggest that prevalence rates are relatively low. Community estimates of male orgasmic disorder range from 0% (Schiavi et al., 1995) to 3% across six studies (Fugl-Meyer & Sjogren Fugl-Meyer, 1999; Lindal & Stefansson, 1993; Singer et al., 1992; Solstad & Hertoft, 1993; Ventegodt, 1998). A much higher 1- year prevalence estimate of 8% is reported by Laumann et al. (1999). Estimates from eight primary care samples across four studies range from 0% (Catalan et al., 1992a) to 36% (El-Rufaie et al., 1997) with a median of 9% (Shahar et al., 1991).
Estimates from four studies in sexuality clinics report current prevalence estimates from 0% (Bhui et al., 1994) to 38% (Catalan et al., 1992b). One study reported a lifetime prevalence of 39% in gay men (Rosser et al., 1997). Estimates from samples of gay men were notably higher than other samples (i.e., ≥ 38% Catalan et al., 1992b; Rosser et al., 1997 vs. ≤ 6% Bhui et al., 1994; Verma et al., 1998). We hypothesize that this difference may reflect greater recognition of the threat of infection with HIV. In this regard, two studies report an increased prevalence of male orgasmic disorder among men with HIV. Current estimates were 20% (vs. 0%) (Catalan et al., 1992a) and 38% (vs. 9%) (Catalan et al., 1992b).
Premature ejaculation is defined as ejaculation with minimal stimulation before the person wishes it. Age, novelty of the sexual situation, and recent frequency of sexual activity are considered in determining premature ejaculation. Community estimates of the current / 1- year prevalence of premature ejaculation range from 4% (Ernst et al., 1993; Fugl-Meyer & Sjogren Fugl-Meyer, 1999) - 5% (Schiavi, 1995; Ventegodt, 1998) across three studies. Two additional studies report significantly higher 1-year estimates of 14% (Solstad et al., 1993) and 29% (Laumann et al., 1999). The reason for these discrepant estimates is not clear. Schiavi et al. (1995) report a current estimate of 20% among former alcohol dependent men.
In primary care settings current estimates range from 2% (Nirenberg et al., 1991) to 31% (Read et al., 1997) across four studies. The lowest estimate was among alcohol dependent individuals choosing not to participate in additional research that would have involved intrusive measurements. Volunteers for the study reported a prevalence of 24% (Nirenberg et al., 1991). Thus, prevalence of premature ejaculation may be estimated at between 4% (Catalan et al., 1992b) and 31% (Read et al., 1997) .
There is an extremely large range of estimates of the current prevalence of premature ejaculation in sexuality clinic samples. Current estimates range from 0% (Bhui et al., 1994; Catalan et al., 1992b) to 77% (Verma et al., 1998) across five studies. The highest estimate (77%; (Verma et al., 1998) is from a Northern Indian population. This figures is nearly four times the next highest estimate of 22% (Goldmeier et al., 1997). If the estimate of Verma et al. (1998) is considered an outlier, a more accurate range is 0% (Bhui et al., 1994; Catalan et al., 1992b) to 22% (Goldmeier et al., 1997). Prevalence of premature ejaculation does not appear to be higher among individuals attending sexuality clinics than in primary care settings.
Erectile Disorder is characterized by inadequate erections for sexual activity. The individual may have difficulty either attaining or maintaining an erection. Current / 1 - year prevalence estimates in the general population range from 0% (Ernst et al., 1993) to 10% (Laumann et al., 1999) across ten studies. The prevalence of erectile disorder increases with age; history of heart disease; diabetes; treated hypertension; untreated ulcer; arthritis; allergy; and smoking (Feldman et al., 1994; Mannino et al., 1994; Panser et al., 1995; Ventegodt, 1998).
Five studies have examined current erectile disorder in older men in the community (Cogen & Steinman, 1990; Feldman et al., 1994; Jonler et al., 1995; Panser et al., 1995; Schiavi et al., 1991). Estimates range from 20% reporting erections less than half the time when sexually stimulated in the last year (Jonler et al., 1995) to 52% (Feldman et al., 1994). The estimate by Feldman et al. (1994) combines “minimal, moderate, and complete” erectile dysfunction. The prevalence of moderate erectile dysfunction in the sample is 25%, closer to that of Jonler et al. (1995).
In general practice settings, current estimates of erectile disorder range from 0.4% (Wei et al., 1994) to 37% (Singer et al., 1992) across seven studies. This wide fluctuation can be attributed to differences among assessment criteria and presence of important risk factors in the samples, including advanced age, medications (Read et al., 1997), diabetes, and medicated hypertension (Modebe, 1990). The highest rates were reported among patients with Parkinson’s disease (60%; Singer et al., 1992) and Alzheimer’s disease (55%; Zeiss et al., 1990).
In sexuality clinics current rates of erectile dysfunction range from 1% (Jindal & Dhall, 1990) to 53% (Bhui et al., 1994) across seven studies. The lowest estimate was based upon interviews of women regarding the sexual functioning of both themselves and their male partner. It is of note that not one of these studies utilized DSM criteria and few provided operational definitions. Thus, it is not surprising that there exists such disparity in prevalence estimates. Goldmeier et al. (1997), Rosser et al. (1997), and Verma et al. (1998) may be the most methodologically sound studies in respect to sample size and criteria. These studies provide current estimates of 19%, 15%, and 24% respectively. Rosser et al. (1997) report lifetime estimates of 40% (getting an erection) and 46% maintaining an erection.
Hypoactive Sexual Desire Disorder is characterized by deficient sexual fantasies and desire for sexual activity. Determination of dysfunction is relative to age and the context of the person’s life. Estimates range from a current / 1-year prevalence of 0% (Schiavi et al., 1995) to 7% (Ernst et al., 1993) across seven community samples. studies. Panser et al. (1995) report a significant positive correlation between age and HSDD (age 70 and over prevalence = 26%). Laumann et al. (1999) report the highest estimate in the general population (1 - year; 16%)
Three studies report prevalence data on HSDD among men in primary care settings. Estimates range from a current prevalence of 3% (Jamieson & Steege, 1996) to 55% (Catalan et al., 1992a) among individuals with HIV. Three studies also examined HSDD in sexuality clinics (Bhui et al., 1994; Catalan et al., 1992b; Rosser et al., 1997). Current estimates of HSDD ranged from 0% (Bhui et al., 1994) to 75% (among HIV+ males; Catalan et al., 1992b).
Three studies examined HSDD in relation to HIV infection status (Catalan et al., 1992a; Catalan et al., 1992b; Pace et al., 1990). Estimates of current HSDD among HIV+ persons range from 13% (Pace et al., 1990) to 75% (Catalan et al., 1992b). The two studies by Catalan et al (1992a, 1992b) report substantially higher estimates than the other studies (75% and 55% respectively). Unfortunately they do not report any criteria and it is unclear to what extent their “loss of interest in sex” corresponds to HSDD. Catalan and colleagues did not find statistically significant differences of prevalence of HSDD across HIV+/- groups. However, Pace et al. (1990) did find a higher prevalence among HIV+ persons than a control sample from an alcohol treatment center.
Dyspareunia is characterized by recurrent genital pain during sexual intercourse. Sexual pain disorder among men appears to be significantly less prevalent than in women (Fass et al., 1998). Estimates across seven studies range from a lifetime prevalence of 0.2% in a random population sample (Lindal & Stefansson, 1993) to a lifetime prevalence of 8% in a combined community and clinical sample (Metz & Seifert, 1990). One study examined painful insertive and receptive anal sex in gay men (Rosser et al., 1997). The study reported current prevalence estimates of 3% (insertive) and 16% (receptive).
Our review of recent prevalence estimates for the sexual dysfunctions is largely consistent with that reported 10 years ago by Spector and Carey (1990). Community samples indicate a current prevalence of 0-3% for male orgasmic disorder, 0-5% for erectile disorder, and 0-3% for male hypoactive sexual desire disorder (HSDD). Pooling current and 1-year figures provides community prevalence estimates of 7-10% for female orgasmic disorder, and 4-5% for premature ejaculation. For the point of comparison, Spector and Carey (1990) reported a current prevalence of 4-10% for male orgasmic disorder, 4-9% for male erectile disorder, 5-10% for female orgasmic disorder, and 36-38% for premature ejaculation. Thus, only the prevalence of premature ejaculation is markedly different. The high estimate for premature ejaculation reported by Spector and Carey (1990) was based upon two relatively small samples. The much lower estimate obtained in this review is based upon four studies with a total of over 2000 men and is thus more representative of the general population. The current review was able to provide an estimate of the prevalence of male HSDD, a figure unavailable previously. Stable community estimates of other sexual dysfunctions remain uncertain.
Spector and Carey (1990) made four suggestions for new research in this area. Specifically, they called for increased use of (1) stratified samples representative of the general population; (2) psychometrically sound assessment techniques to facilitate interpretation and replication; (3) a common classification system to aid comparison across studies; and (4) collection of incidence data. There are some notable studies over the past ten years that have incorporated these methodological recommendations. For example, Ernst et al. (1993) and Rekers et al. (1992) used stronger sampling techniques, stratifying by psychological distress and age, respectively. The field has also had the benefit of several larger scale (> 1000 participants) random population surveys (e.g., Barlow et al., 1997; Fugl-Meyer & Sjogren Fugl-Meyer, 1999; Laumann et al., 1999; Ventegodt, 1998).
There has also been progress in assessment techniques. For example, Brown et al. (1990) and Goggin et al. (1998) use a modified version of the Structured Clinical Interview for DSM-III-R (Brown & Rundell, 1993). Reports of inter-rater reliability in several studies provide a measure of reliability of diagnoses (Brown & Rundell, 1990; Meyer-Bahlburg et al., 1993). Additional assessment instruments with known psychometric characteristics are also being used (e.g., the DISS-IIIA was used by Robins, 1986; Chandraiah et al., 1991; Lindal & Stefansson, 1993, and Meyer-Bahlburg et al., 1993; the GRISS was used by Rust & Golombok, 1986 and Goldmeier et al., 1997). Review of the studies in which the most psychometrically sound assessment techniques were used also demonstrates a trend toward using the DSM as a common classification system.
Incidence data continues to be sparse. Wei et al. (1994) is one exception. These authors report incidence data on erectile dysfunction stratified by age.
There have been a small number of excellent studies that have incorporated many important methodological features into study design (e.g., Fugl-Meyer & Sjogren Fugl-Meyer, 1999; Lindal et al., 1993). However, despite the increased attention in the past decade to the study of sexual dysfunctions there appears to have been relatively little methodological improvement overall. We identify three successive strategies for improvement in relation to assessment criteria. First, the criteria for determining a dysfunction need to be clearly reported. Although several investigators have used operational definitions, many studies failed to report the criteria they used in the paper. Lack of consistent reporting of assessment criteria make comparisons across studies difficult and hinder the accumulation of data across studies to enhance knowledge. Second, standard criteria for the sexual dysfunctions need to be adopted. The use of standard diagnostic criteria appears to be the exception rather than the rule among the studies reviewed. The DSM and the multiaxial system proposed by Schover et al. (1982) provide two potential options.
The third avenue for development is to examine sexual disorders rather than simply the dysfunction. The omission of psychological sequelae of the sexual dysfunctions is a significant methodological concern. According to the DSM-IV (APA, 1994), sexual disorder diagnoses need to be based on three criteria: (A) sexual dysfunction (i.e., physical / psychological manifestation (e.g., lack of orgasm, lack of erection, pain during intercourse, lack of sexual interest, etc.)), (B) the disturbance causes marked distress or interpersonal difficulty, and (C) the dysfunction is not better accounted for by another Axis I disorder (except another sexual dysfunction) and is not due to the physiological effects of a substance or general medical condition. Criteria A has frequently been incorporated into most investigators’ operational definitions. However, criteria B and C are typically omitted from the diagnostic criteria. From a clinical standpoint, the report of accompanying distress and/or interpersonal difficulty is important. However, this criterion was rarely addressed in the reported criteria in the studies reviewed. Typically, only the prevalence of symptoms is reported although some exceptions can be noted. For example, Amr, Halim, and Moussa (1997) report the prevalence of both erectile disorder determined by DSM-III-R criteria as well as prevalence of erectile dysfunction symptoms. The latter led to a prevalence rate of 27% whereas the former resulted in a much lower (5%) rate. This additional level of detail provides especially helpful information regarding the underlying development of the disorder. For example, Amr et al. (1997) reported significant increases in erectile dysfunction symptoms but not erectile disorder in relation to pesticide exposure. Such findings may be informative in understanding the biological and psychological contributions to the development of sexual disorders.
The study by Fugl-Meyer and Sjogren Fugl-Meyer (1999) is particularly informative in respect to the relations between sexual dysfunction and sexual disorder characterized by resultant perceived psychosocial problems. This study assessed the prevalence of the dysfunctions as well as the percentage of participants who perceived their sexual dysfunction as problematic. For some dysfunctions, there was a high concordance between the presence of a dysfunction and perceived problems. For example, sixty-nine percent of the men reporting erectile dysfunction reported that it was problematic. In contrast, only forty-five percent of women with orgasmic dysfunction perceived it as problematic. Thus, in this study, if one defined female orgasmic disorder as the inability to attain orgasm the 1-year prevalence rate is 22%. In contrast, the prevalence rate is only 10% if one defines the disorder as the presence of the dysfunction and the dysfunction causing a problem (marked distress or interpersonal difficulty in DSM-IV terminology). This study clearly differentiates between sexual functioning on the one hand and a psychological disorder defined in part by subjective distress and disturbance in interpersonal relations. This study provides a clear demonstration of how differences in diagnostic criteria can have profound a effect on prevalence estimates. Such differences contribute to the wide discrepancies seen across some studies.
We acknowledge that most studies were designed only to obtain data on the occurrence of a symptom and that investigators did not claim to be assessing a disorder defined in the DSM. Determination of the appropriateness of assessing a sexual dysfunction versus a disorder (in the DSM sense) rests upon the goals of the study. Assessing solely the dysfunction is appropriate if the potential accompanying distress or interpersonal conflict is not of interest. In some cases, the symptom is an important focal point as in the relationship between erectile dysfunction and health problems such as diabetes mellitus (Weinhardt & Carey, 1996). In such studies, biological precursors are of interest. In more clinically focused research, determination of whether or not the dysfunction is accompanied by significant distress or interpersonal conflict is relevant. It is such psychosocial problems that are the impetus for intervention not variation in sexual functioning per se. For symptoms such as reduced sex drive, the importance of the symptom in isolation from DSM criteria B or C is unknown, and the prevalence estimates are less useful.
The wealth of studies conducted over the past ten years is encouraging as are the adoption of the methodological suggestions that were outlined in Spector and Carey (1990). We note above some of the studies that have particularly sound methodological designs. These studies are exemplars that could guide the continuing development in the study of sexual functioning. These exemplars are unfortunately few in comparison to the full collection of studies. Many continue to have methodological problems that limit their potential usefulness. With continued attention to statistical design, it is hoped that methodologically rigorous studies will no longer be the exception to the rule.
- Aizenberg D, Zemishlany Z, Dorfman-Etrog P, Weizman A. Sexual dysfunction in male schizophrenic patients. J. Clin. Psychiatry. 1995;56:137–141.[PubMed]
- American Psychiatric Association . Diagnostic and statistical manual of mental disorders. 4th ed. Author; Washington, D.C.: 1994.
- Amr MM, Halim ZS, Moussa SS. Psychiatric disorders among Egyptian pesticide applicators and formulators. Environ. Res. 1997;73:193–199.[PubMed]
- Barlow DH, Cardozo LD, Francis RM, Griffin M, Hart DM, Stephens E, Sturdee DW. Urogenital aging and its effect on sexual health in older British women. Br. J. Obs. Gyn. 1997;104:87–91.[PubMed]
- Bhui K, Herriot P, Dein S, Watson JP. Asians presenting to a sex and marital therapy clinic. Int. J. Soc. Psychiatry. 1994;40:194–204.[PubMed]
- Boulet MJ, Oddens BJ, Lehert P, Vemer HM, Visser A. Climacteric and menopause in seven South-east Asian countries. Maturitas. 1994;19:157–176.[PubMed]
- Brown GR, Rundell JB. A prospective study of psychiatric aspects of early HIV disease in women. Gen. Hosp. Psychiatry. 1993;15:139–147.[PubMed]
- Brown GR, Rundell JR. Prospective study of psychiatric morbidity in HIV-seropositive women without AIDS. Gen. Hosp. Psychiatry. 1990;12:30–35.[PubMed]
- Catalan J, Klimes I, Bond A, Day A, Garrod A, Rizza C. The psychosocial impact of HIV infection in men with haemophilia: controlled investigation and factors associated with psychiatric morbidity. J. Psychosom. Res. 1992a;36:409–416.[PubMed]
- Catalan J, Klimes I, Day A, Garrod A, Bond A, Gallwey J. The psychosocial impact of HIV infection in gay men. A controlled investigation and factors associated with psychiatric morbidity. Br. J. Psychiatry. 1992b;161:774–778.[PubMed]
- Chandraiah S, Levenson JL, Collins JB. Sexual dysfunction, social maladjustment, and psychiatric disorders in women seeking treatment in a premenstrual syndrome clinic. Int. J. Psychiatry Med. 1991;21:189–204.[PubMed]
- Chiechi LM, Granieri M, Lobascio A, Ferreri R, Loizzi P. Sexuality in the climacterium. Clin. Exp. Obstet. Gynecol. 1997;24:158–159.[PubMed]
- Cogen R, Steinman W. Sexual function and practice in elderly men of lower socioeconomic status. J. Fam. Pract. 1990;31:162–166.[PubMed]
- Derogatis LR. SCL-90. Administration, scoring, and procedures for the R (revised) version and other instruments of the psychopathology rating scale series. Johns Hopkins Press; Baltimore: 1977.
- Derogatis LR, Meyer J. A psychological profile of the sexual dysfunctions. Arch. Sex. Behav. 1979;8:201–223.[PubMed]
- Diokno AC, Brown MB, Herzog AR. Sexual function in the elderly. Arch. Intern. Med. 1990;150:197–200.[PubMed]
- El-Rufaie OEF, Bener A, Abuzeid MSO, Ali TA. Sexual dysfunction among type II diabetic men: A controlled study. J. Psychosom. Res. 1997;43:605–612.[PubMed]
- Ernst C, Foldenyi M, Angst J. The Zurich Study: XXI. Sexual dysfunctions and disturbances in young adults. Data of a longitudinal epidemiological study. Eur. Arch. Psychiatry Clin. Neurosci. 1993;243:179–188.[PubMed]
- Fass R, Fullerton S, Naliboff B, Hirsh T, Mayer EA. Sexual dysfunction in patients with irritable bowel syndrome and non- ulcer dyspepsia. Digestion. 1998;59:79–85.[PubMed]
- Feldman HA, Goldstein I, Hatzichristou DG, Krane RJ, McKinlay JB. Impotence and its medical and psychosocial correlates: results of the Massachusetts Male Aging Study. J. Urol. 1994;151:54–61.[PubMed]
- Fugl-Meyer AR, Sjogren Fugl-Meyer K. Sexual disabilities, problems, and satisfaction in 18-74 year old Swedes. Scand J Sexol. 1999;3:79–105.
- Glatt AE, Zinner SH, McCormack WM. The prevalence of dyspareunia. Obstet. Gynecol. 1990;75:433–436.[PubMed]
- Goggin K, Engelson ES, Rabkin JG, Kotler DP. The relationship of mood, endocrine, and sexual disorders in human immunodeficiency virus positive (HIV+) women: An exploratory study. Psychosom. Med. 1998;60:11–16.[PubMed]
- Goldberg DP, Hillier A. Manual of the general health questionnaire. Egyptian Psychiatric Association: Cairo: 1979.
- Goldmeier D, Keane FE, Carter P, Hessman A, Harris JR, Renton A. Prevalence of sexual dysfunction in heterosexual patients attending a central London genitourinary medicine clinic. Int. J. STD AIDS. 1997;8:303–306.[PubMed]
- Heisterberg L. Factors influencing spontaneous abortion, dyspareunia, dysmenorrhea, and pelvic pain. Obstet. Gynecol. 1993;81:594–597.[PubMed]
- Ismael NN. A study on the menopause in Malaysia. Maturitas. 1994;19:205–209.[PubMed]
- Jamieson DJ, Steege JF. The prevalence of dysmenorrhea, dyspareunia, pelvic pain, and irritable bowel syndrome in primary care practices. Obstet. Gynecol. 1996;87:55–58.[PubMed]
- Jindal UN, Dhall GI. Psychosexual problems of infertile women in India. Int J Fertil. 1990;35:222–225.[PubMed]
- Jonler M, Moon T, Brannan W, Stone NN, Heisey D, Bruskewitz RC. The effect of age, ethnicity and geographical location on impotence and quality of life. Br. J. Urol. 1995;75:651–655.[PubMed]
- Laumann EO, Paik A, Rosen RC. Sexual Dysfunction in the United States. JAMA. 1999;281:537–544.[PubMed]
- Lindal E, Stefansson JG. The lifetime prevalence of psychosexual dysfunction among 55 to 57-year-olds in Iceland. Soc. Psychiatry Psychiatr. Epidemiol. 1993;28:91–95.[PubMed]
- Mannino DM, Klevens RM, Flanders WD. Cigarette smoking: an independent risk factor for impotence? Am. J. Epidemiol. 1994;140:1003–1008.[PubMed]
- Metz M, Seifert MH. Men’s expectations of physicians in sexual health concerns. J. Sex Marital Ther. 1990;16:79–88.[PubMed]
- Meyer-Bahlburg HF, Nostlinger C, Exner TM, Ehrhardt AA, Gruen RS, Lorenz G, Gorman JM, el-Sadr W, Sorrell SJ. Sexual functioning in HIV+ and HIV- injected drug-using women. J. Sex Marital Ther. 1993;19:56–68.[PubMed]
- Modebe O. Erectile failure among medical clinic patients. Afr J Med Med Sci. 1990;19:259–264.[PubMed]
- Moody GA, Mayberry JF. Perceived sexual dysfunction amongst patients with inflammatory bowel disease. Digestion. 1993;54:256–260.[PubMed]
- Nirenberg TD, Wincze JP, Bansal S, Liepman MR, et al. Volunteer bias in a study of male alcoholics’ sexual behavior. Arch. Sex. Behav. 1991;20:371–379.[PubMed]
- Pace J, Brown GR, Rundell JR, Paolucci S. Prevalence of psychiatric disorders in a mandatory screening program for infection with human immunodeficiency virus: A pilot study. Military Medicine. 1990;155:76–80.[PubMed]
- Panser LA, Rhodes T, Girman CJ, Guess HA, Chute CG, Oesterling JE, Lieber MM, Jacobsen SJ. Sexual function of men ages 40 to 79 years: the Olmsted County Study of Urinary Symptoms and Health Status Among Men. J. Am. Geriatr. Soc. 1995;43:1107–1111.[PubMed]
- Ramoso-Jalbuena J. Climacteric Filipino women: a preliminary survey in the Philippines. Maturitas. 1994;19:183–190.[PubMed]
- Read S, King M, Watson J. Sexual dysfunction in primary medical care: prevalence, characteristics and detection by the general practitioner. J. Public Health Med. 1997;19:387–391.[PubMed]
- Rekers H, Drogendijk AC, Valkenburg HA, Riphagen F. The menopause, urinary incontinence and other symptoms of the genito-urinary tract. Maturitas. 1992;15:101–111.[PubMed]
- Reynolds CF, Frank E, Thase ME, Houck PR, Jennings JR, Howell JR, Lilienfeld SO, Kupfer DJ. Assessment of sexual function in depressed, impotent, and healthy men: Factor analysis of a brief sexual function questionnaire for men. Psychiatry Res. 1988;24:231–250.[PubMed]
- Robins LN, Helzer JE, Croughan J. The NIMH Diagnostic Interview Schedule: Version III. Public Health Service (HSS); 1981.
- Rosser BR, Metz ME, Bockting WO, Buroker T. Sexual difficulties, concerns, and satisfaction in homosexual men: an empirical study with implications for HIV prevention. J. Sex Marital Ther. 1997;23:61–73.[PubMed]
- Rust J, Golombok S. The Golombok-Rust Inventory of Sexual Satisfaction (GRISS) NFER/Nelson; Windsor: 1986.
- Schiavi RC, Mandeli J, Schreiner-Engel P, Chambers A. Aging, sleep disorders, and male sexual function. Biol. Psychiatry. 1991;30:15–24.[PubMed]
- Schiavi RC, Stimmel BB, Mandeli J, White D. Chronic alcoholism and male sexual function. Am. J. Psychiatry. 1995;152:1045–1051.[PubMed]
- Schover LR, Friedman JM, Weiler SJ, Heiman JR, LoPiccolo J. Multiaxial problem-oriented system for sexual dysfunctions: an alternative to DSM-III. Arch. Gen. Psychiatry. 1982;39:614–619.[PubMed]
- Shahar E, Lederer J, Herz MJ. The use of a self-report questionnaire to assess the frequency of sexual dysfunction in family practice clinics. Fam. Pract. 1991;8:206–212.[PubMed]
- Singer C, Weiner WJ, Sanchez-Ramos JR. Autonomic dysfunction in men with Parkinson’s disease. Eur. Neurol. 1992;32:134–140.[PubMed]
- Solstad K, Hertoft P. Frequency of sexual problems and sexual dysfunction in middle-aged Danish men. Arch. Sex. Behav. 1993;22:51–58.[PubMed]
- Spector IP, Carey MP. Incidence and prevalence of the sexual dysfunctions: A critical review of the empirical literature. Arch. Sex. Behav. 1990;19(4):389–408.[PubMed]
- Ventegodt S. Sex and the quality of life in Denmark. Arch. Sex. Behav. 1998;27:295–307.[PubMed]
- Verma KK, Khaitan BK, Singh OP. The frequency of sexual dysfunctions in patients attending a sex therapy clinic in north India. Arch. Sex. Behav. 1998;27:309–314.[PubMed]
- Wasti S, Robinson SC, Akhtar Y, Khan S, Badaruddin N. Characteristics of menopause in three socioeconomic urban groups in Karachi, Pakistan. Maturitas. 1993;16:61–69.[PubMed]
- Weber AM, Walters MD, Schover LR, Mitchinson A. Vaginal anatomy and sexual function. Obstet. Gynecol. 1995;86:946–949.[PubMed]
- Wei M, Macera CA, Davis DR, Hornung CA, Nankin HR, Blair SN. Total Cholesterol and High Density Lipoprotein Cholesterol as Important Predictors of Erectile Dysfunction. Am. J. Epidemiol. 1994;140:930–937.[PubMed]
- Weinhardt LS, Carey MP. Prevalence of erectile disorder among men with diabetes mellitus: Review of the empirical literature. J. Sex Res. 1996;33:205–214.
- Whelan EA, Grajewski B, Wild DK, Schnorr TM, Alderfer R. Evaluation of reproductive function among men occupationally exposed to a stilbene derivative: II. perceived libido and potency. Am J Ind Med. 1996;29:59–65.[PubMed]
- Zeiss AM, Davies HD, Wood M, Tinklenberg JR. The incidence and correlates of erectile problems in patients with Alzheimer’s disease. Arch. Sex. Behav. 1990;19:325–331.[PubMed]
The present paper aims to provide basic guidelines to present epidemiological data using tables and graphs in Dermatology. Although simple, the preparation of tables and graphs should follow basic recommendations, which make it much easier to understand the data under analysis and to promote accurate communication in science. Additionally, this paper deals with other basic concepts in epidemiology, such as variable, observation, and data, which are useful both in the exchange of information between researchers and in the planning and conception of a research project.
Keywords: Epidemiology, Epidemiology, descriptive, Tables
Among the essential stages of epidemiological research, one of the most important is the identification of data with which the researcher is working, as well as a clear and synthetic description of these data using graphs and tables. The identification of the type of data has an impact on the different stages of the research process, encompassing the research planning and the production/publication of its results. For example, the use of a certain type of data impacts the amount of time it will take to collect the desired information (throughout the field work) and the selection of the most appropriate statistical tests for data analysis.
On the other hand, the preparation of tables and graphs is a crucial tool in the analysis and production/publication of results, given that it organizes the collected information in a clear and summarized fashion. The correct preparation of tables allows researchers to present information about tens or hundreds of individuals efficiently and with significant visual appeal, making the results more easily understandable and thus more attractive to the users of the produced information. Therefore, it is very important for the authors of scientific articles to master the preparation of tables and graphs, which requires previous knowledge of data characteristics and the ability of identifying which type of table or graph is the most appropriate for the situation of interest.
Before evaluating the different types of data that permeate an epidemiological study, it is worth discussing about some key concepts (herein named data, variables and observations):
Data - during field work, researchers collect information by means of questions, systematic observations, and imaging or laboratory tests. All this gathered information represents the data of the research. For example, it is possible to determine the color of an individual's skin according to Fitzpatrick classification or quantify the number of times a person uses sunscreen during summer.1,2 All the information collected during research is generically named "data." A set of individual data makes it possible to perform statistical analysis. If the quality of data is good, i.e., if the way information was gathered was appropriate, the next stages of database preparation, which will set the ground for analysis and presentation of results, will be properly conducted.
Observations - are measurements carried out in one or more individuals, based on one or more variables. For instance, if one is working with the variable "sex" in a sample of 20 individuals and knows the exact amount of men and women in this sample (10 for each group), it can be said that this variable has 20 observations.
Variables - are constituted by data. For instance, an individual may be male or female. In this case, there are 10 observations for each sex, but "sex" is the variable that is referred to as a whole. Another example of variable is "age" in complete years, in which observations are the values 1 year, 2 years, 3 years, and so forth. In other words, variables are characteristics or attributes that can be measured, assuming different values, such as sex, skin type, eye color, age of the individuals under study, laboratory results, or the presence of a given lesion/disease. Variables are specifically divided into two large groups: (a) the group of categorical or qualitative variables, which is subdivided into dichotomous, nominal and ordinal variables; and (b) the group of numerical or quantitative variables, which is subdivided into continuous and discrete variables.
Dichotomous variables, also known as binary variables: are those that have only two categories, i.e., only two response options. Typical examples of this type of variable are sex (male and female) and presence of skin cancer (yes or no).
Ordinal variables: are those that have three or more categories with an obvious ordering of the categories (whether in an ascending or descending order). For example, Fitzpatrick skin classification into types I, II, III, IV and V.1
Nominal variables: are those that have three or more categories with no apparent ordering of the categories. Example: blood types A, B, AB, and O, or brown, blue or green eye colors.
Discrete variables: are observations that can only take certain numerical values. An example of this type of variable is subjects' age, when assessed in complete years of life (1 year, 2 years, 3 years, 4 years, etc.) and the number of times a set of patients visited the dermatologist in a year.
Continuous variables: are those measured on a continuous scale, i.e., which have as many decimal places as the measuring instrument can record. For instance: blood pressure, birth weight, height, or even age, when measured on a continuous scale.
It is important to point out that, depending on the objectives of the study, data may be collected as discrete or continuous variables and be subsequently transformed into categorical variables to suit the purpose of the research and/or make interpretation easier. However, it is important to emphasize that variables measured on a numerical scale (whether discrete or continuous) are richer in information and should be preferred for statistical analyses. Figure 1 shows a diagram that makes it easier to understand, identify and classify the abovementioned variables.
DATA PRESENTATION IN TABLES AND GRAPHS
Firstly, it is worth emphasizing that every table or graph should be self-explanatory, i.e., should be understandable without the need to read the text that refers to it refers.
Presentation of categorical variables
In order to analyze the distribution of a variable, data should be organized according to the occurrence of different results in each category. As for categorical variables, frequency distributions may be presented in a table or a graph, including bar charts and pie or sector charts. The term frequency distribution has a specific meaning, referring to the the way observations of a given variable behave in terms of its absolute, relative or cumulative frequencies.
In order to synthesize information contained in a categorical variable using a table, it is important to count the number of observations in each category of the variable, thus obtaining its absolute frequencies. However, in addition to absolute frequencies, it is worth presenting its percentage values, also known as relative frequencies. For example, table 1 expresses, in absolute and relative terms, the frequency of acne scars in 18-year-old youngsters from a population-based study conducted in the city of Pelotas, Southern Brazil, in 2010.3
Absolute and relative frequencies of acne scar in 18- year-old adolescents (n = 2.414). Pelotas, Brazil, 2010
The same information from table 1 may be presented as a bar or a pie chart, which can be prepared considering the absolute or relative frequency of the categories. Figures 2 and 3 illustrate the same information shown in table 1, but present it as a bar chart and a pie chart, respectively. It can be observed that, regardless of the form of presentation, the total number of observations must be mentioned, whether in the title or as part of the table or figure. Additionally, appropriate legends should always be included, allowing for the proper identification of each of the categories of the variable and including the type of information provided (absolute and/or relative frequency).
Absolute frequencies of acne scar in 18-year-old adolescents (n = 2.414). Pelotas, Brazil, 2010
Relative frequencies of acne scar in 18-year-old adolescents (n = 2.414). Pelotas, Brazil, 2010
Presentation of numerical variables
Frequency distributions of numerical variables can be displayed in a table, a histogram chart, or a frequency polygon chart. With regard to discrete variables, it is possible to present the number of observations according to the different values found in the study, as illustrated in table 2. This type of table may provide a wide range of information on the collected data.
Educational level of 18-year-old adolescents (n = 2,199). Pelotas, Brazil, 2010
Table 2 shows the distribution of educational levels among 18-year-old youngsters from Pelotas, Southern Brazil, with absolute, relative, and cumulative relative frequencies. In this case, absolute and relative frequencies correspond to the absolute number and the percentage of individuals according to their distribution for this variable, respectively, based on complete years of education. It should be noticed that there are 450 adolescents with 8 years of education, which corresponds to 20.5% of the subjects. Tables may also present the cumulative relative frequency of the variable. In this case, it was found that 50.6% of study subjects have up to 8 years of education. It is important to point that, although the same data were used, each form of presentation (absolute, relative or cumulative frequency) provides different information and may be used to understand frequency distribution from different perspectives.
When one wants to evaluate the frequency distribution of continuous variables using tables or graphs, it is necessary to transform the variable into categories, preferably creating categories with the same size (or the same amplitude). However, in addition to this general recommendation, other basic guidelines should be followed, such as: (1) subtracting the highest from the lowest value for the variable of interest; (2) dividing the result of this subtraction by the number of categories to be created (usually from three to ten); and (3) defining category intervals based on this last result.
For example, in order to categorize height (in meters) of a set of individuals, the first step is to identify the tallest and the shortest individual of the sample. Let us assume that the tallest individual is 1.85m tall and the shortest, 1.55m tall, with a difference of 0.3m between these values. The next step is to divide this difference by the number of categories to be created, e.g., five. Thus, 0.3m divided by five equals 0.06m, which means that categories will have exactly this range and will be numerically represented by the following range of values: 1st category - 1.55m to 1.60m; 2nd category - 1.61m to 1.66m; 3rd category - 1.67m to 1.72m; 4th category - 1.73m to 1.78m; 5th category - 1.79m to 1.85m.
Table 3 illustrates weight values at 18 years of age in kg (continuous numerical variable) obtained in a study with youngsters from Pelotas, Southern Brazil.4,5Figure 4 shows a histogram with the variable weight categorized into 20-kg intervals. Therefore, it is possible to observe that data from continuous numerical variables may be presented in tables or graphs.
Weight distribution among 18-year-old young male sex (n = 2.194). Pelotas, Brazil, 2010
Weight distribution at 18 years of age among youngsters from the city of Pelotas. Pelotas (n = 2.194), Brazil, 2010
Assessing the relationship between two variables
The forms of data presentation that have been described up to this point illustrated the distribution of a given variable, whether categorical or numerical. In addition, it is possible to present the relationship between two variables of interest, either categorical or numerical.
The relationship between categorical variables may be investigated using a contingency table, which has the purpose of analyzing the association between two or more variables. The lines of this type of table usually display the exposure variable (independent variable), and the columns, the outcome variable (dependent variable). For example, in order to study the effect of sun exposure (exposure variable) on the development of skin cancer (outcome variable), it is possible to place the variable sun exposure on the lines and the variable skin cancer on the columns of a contingency table. Tables may be easier to understand by including total values in lines and columns. These values should agree with the sum of the lines and/or columns, as appropriate, whereas relative values should be in accordance with the exposure variable, i.e., the sum of the values mentioned in the lines should total 100%.
It is such a display of percentage values that will make it possible for risk or exposure groups to be compared with each other, in order to investigate whether individuals exposed to a given risk factor show higher frequency of the disease of interest. Thus, table 4 shows that 75.0%, 9.0%, and 0.3% of individuals in the study sample who had been working exposed to the sun for 20 years or more, for less than 20 years, and had never been working exposed to the sun, respectively, developed non-melanoma skin cancer. Another way of interpreting this table is observing that 25.0%, 91%,.0%, and 99.7% of individuals who had been working exposed to the sun for 20 years of more, for less than 20 years, and had never been working exposed to the sun did not develop non-melanoma skin cancer. This form of presentation is one of the most used in the literature and makes the table easier to read.
Sun exposure during work and non-melanoma skin cancer (hypothetical data).
The relationship between two numerical variables or between one numerical variable and one categorical variable may be assessed using a scatter diagram, also known as dispersion diagram. In this diagram, each pair of values is represented by a symbol or a dot, whose horizontal and vertical positions are determined by the value of the first and second variables, respectively. By convention, vertical and horizontal axes should correspond to outcome and exposure variables, respectively. Figure 5 shows the relationship between weight and height among 18-year-old youngsters from Pelotas, Southern Brazil, in 2010.3,4 The diagram presented in figure 5 should be interpreted as follows: the increase in subjects' height is accompanied by an increase in their weight.
Point diagram for the relationship between weight (kg) and height (cm) among 18-year-old youngsters from the city of Pelotas (n = 2.194). Pelotas, Brazil, 2010.
BASIC RULES FOR THE PREPARATION OF TABLES AND GRAPHS
Ideally, every table should:
Present values with the same number of decimal places in all its cells (standardization);
Include a title informing what is being described and where, as well as the number of observations (N) and when data were collected;
Have a structure formed by three horizontal lines, defining table heading and the end of the table at its lower border;
Not have vertical lines at its lateral borders;
Provide additional information in table footer, when needed;
Be inserted into a document only after being mentioned in the text; and
Be numbered by Arabic numerals.
Similarly to tables, graphs should:
Include, below the figure, a title providing all relevant information;
Be referred to as figures in the text;
Identify figure axes by the variables under analysis;
Quote the source which provided the data, if required;
Demonstrate the scale being used; and
The graph's vertical axis should always start with zero. A usual type of distortion is starting this axis with values higher than zero. Whenever it happens, differences between variables are overestimated, as can been seen in figure 6.
Figure showing how graphs in which the Y-axis does not start with zero tend to overestimate the differences under analysis. On the left there is a graph whose Y axis does not start with zero and on the right a graph reproducing the same data but with...
Understanding how to classify the different types of variables and how to present them in tables or graphs is an essential stage for epidemiological research in all areas of knowledge, including Dermatology. Mastering this topic collaborates to synthesize research results and prevents the misuse or overuse of tables and figures in scientific papers.
Conflict of Interest: None
Financial Support: None
How to cite this article: Duquia RP, Bastos JL, Bonamigo RR, González-Chica DA, Martínez-Mesa J. Presenting data in tables and charts. An Bras Dermatol. 2014;89(2):280-5.
*Work performed at the Dermatology service, Universidade Federal de Ciências da Saúde de Porto Alegre (UFCSPA), Departamento de Saúde Pública e Departamento de Nutrição da UFSC.
1. Walker SL, Hawk JLM, Young AR. Acute and chronic effects. In: Freedberg IM, Eisen AZ, Wolff K, Austen KF, Goldsmith LA, Katz SI, editors. Fitzpatrick's Dermatology in General Medicine. 8th ed. pp. 1275–1281.
2. Duquia RP, Baptista Menezes AM, Reichert FF, de Almeida HL., Jr. Prevalence and associated factors with sunscreen use in Southern Brazil: A population-based study. J Am Acad Dermatol. 2007;57:73–80.[PubMed]
3. Duquia RP, de Almeida HL, Jr., Breunig JA, Souzat PR, Göellner CD. Most common patterns of acne in male adolescents: a population-based study. Int J Dermatol. 2013;52:550–553.[PubMed]
4. Breunig Jde A, de Almeida HL, Jr., Duquia RP, Souza PR, Staub HL. Scalp seborrheic dermatitis: prevalence and associated factors in male adolescents. Int J Dermatol. 2012;51:46–49.[PubMed]
5. Almeida H, Jr., Cecconi J, Duquia RP, Souza PR, Breunig J. Sensitivity and specificity of self-reported acne in 18-year-old adolescent males. Int J Dermatol. 2013;52:946–948.[PubMed]