A comparative analysis of report

A comparative analysis of reports on the prevalence of transsexualism [draft in progress]

In this report, we present a series of cross-comparisons and analyses of past studies of the prevalence of transsexualism, using the notation and methodology for the prevalence calculations presented by Olyslager and Conway [Olyslager and Conway 07] at WPATH 07:

Olyslager, F. and Conway, L., “On the Calculation of the Prevalence of Transsexualism”,

presented at the WPATH 20th International Symposium, Chicago, Illinois, September 6, 2007

The Olyslager and Conway paper identifies critical factors leading to an incommensurability in the results presented in earlier papers, revealing that those earlier results cannot be directly cross-compared in any meaningful way.

However, by taking the incommensurability factors into account, we can reprocess the original data from the earlier papers and derive a series of commensurable results that are directly comparable across the entire set of studies.

Table 1 presents derived commensurable results in "1:N" form (tabulating values of prevalence and incidence as "1 in N") for all major past studies of prevalence. Table 2 presents the same results in the form of "n out of 100,000".

In both tables the relevant studies are down the leftmost column (which contains links to PDF's of those papers for convenient reference). Commensurable derived values for the various types of prevalence and incidence (as a function of the Olyslager-Conway parameters: active, inherent; mf, fm; TS, SH, HT, SRS) are then presented under the corresponding headings in the other columns.

Columns are included in the tables to identify the nationality of each study, and record whether a study portrayed its results as lower bounds (Y) or absolute values (N). The "doc" column links to an annotated set of notes and calculations about each study, while the "note" column links to notes below. Those notes and calculations were compiled during the research into analyses of earlier reports that led up to Olyslager-Conway report, and are in process of being updated now.

Note that Table 1 and Table 2 are also posted online as a spreadsheet, at the following page:

[1]

This is the classic first systematic study on TS prevalence.

Counted numbers of fm and mf approaching psychiatrists in Sweden for sex reassignment up to 1968 to get a lower bound on prevalence.

Defined transsexualism as those seeking SRS.

Set pattern for many later studies.

Indicated that prevalent esitimates are difficult (i.e., are lower bounds due to underreporting).

[2]

Incidence per population of mf and fm name changes in Sweden circa 1970.
Some unknown fraction will have had SRS.

Ip = 0.2 per 100,000 over 15 years of age for both mf and fm

[3]

Uses the defs and methods of Walinder68.

They count numbers of mf and fm approaching their clinic in Manchester area of the UK from 1958-1968. Report that clinic was widely known in that area, and most psychs would have referred such clients to them.

Results coincidentally fall almost on top of [Walinder68], giving impetus to the notion that these levels of prevalence are "inherent" and "representative.

[4]

This paper's results are NOT PREVALENCES

Instead it reports the sums of annual population INCIDENCE over TWO YEARS, and divides by the relevant population and calls the results "prevalence".

It just happens to turn out that by coincidence these "two year incidence values" are fairly close to the earlier work of Walinder in SW and Eklund in NL. Maybe that's why these authors jumped to the conclusion that these WERE prevalence values?

In order to normalize these results and make them comparable with other work, we divide by two to get the annual incidences, and then identify them as annual incidences.

If the authors had done this, they might have noticed that their numbers for Incidence were way higher than priior reports.

[5]

This paper uses Walinder's definition and method, but actually counted those who were at least on hormones already. Thus we need to do a normalization to compare results with Walinder and earlier papers.

This report counted numbers of mf and fm trans people who approached the AZVU clinic in Amsterdam and who received hormone treatment there - and compiled details of the numbers by age groups for the years 1975-1986.

It then focussed in on the years 1980, 83, 86, and compiled annual incidences as a function of age (calling them "prevalence rates") in Table 1 in the report.

It then gave prevalences for the NL in 80, 83, 86, but did not notice that the explanation for the very rapid increase in prevalence was due to the initial transient in available access to treatment (which can be seen in it's Fig 1).

[6]

This paper uses the same criteria as Walinder for identifying transsexual people.

However, it actually counted records of SRS's from the Dept of Obst and Gyn at the National Univ., plus a few reports from two private surgeons.

Thus this paper is reporting on the active Prevalence of SRS, rather than the prevalence of transexualism.

Tsoi88 reports them occurring at an average age of 24.1 (in a population having a life expectancy of ~ 80). By using Equation 5b in [Olylager and Conway 07] we can infer a likely lower bound on P(SRSImf), i.e., the prevalence of those who’ve had mf SRS plus those tracking towards SRS, as follows :

P(SRSImf) derived from [Tsoi88] ≈ 1:2,900 (80 / (80 – 24.1) = 1:2,000

[7]

This report builds on Eklund88, working to bring it up to date.

Its numbers of 1:11,900 and 1: 30,400 have been widely used in recent years (the 00's) by HBIGA/WPATH.

The report uses Walinder's definition and method of counting (approaches to official programs for help).

However, the report reduces the count to include only those who HAVE begun hormone therapy (not just those seeking treatment). Thus the above prevalence numbers are for P(HTAmf) and P(HTAfm), as normalized in the spreadsheet entry for this report.

Furthermore, the abstract hints that the numbers actually represent those who underwent SRS (which is inconsistent with the contents of the report), setting the stage for confusions in later work.

The paper does not mention that prevalence is hard to come by and that any numbers are merely lower bounds. Instead it begins a trend towards thinking of and reporting out the numbers as absolute ones.

[8]

A comparison of the more recent 1972-1992 annuals counts of trans people seeking sex reassignment in Sweden with earlier results of Walinder.

Concludes that incidence has remained "constant".

Revises raw data downward by dismissing any non-"true-transsexuals", etc.

[9]

This report builds on Bakker1993. It provides details on whether the trans people counted were only/mostly on HT or whether they mostly had had SRS.

The report enables us to disambiguate P(HT) vs P(SRS) results in Bakker93:
P(TSmf) = normalized P(HTmf) = 1:11,900 = 8.4 per 100,000
P(TSfm) = normalized P(HTmf) = 1:30,400 = 3.3 per 100,000

We can calculate from data in this 96 report that
Normalized P(SRSmf) = 0.68 P(HTmf) = 1:17,500 = 5.7 per 100,000
Normalized P(SRSmf) = 0.74 P(HTmf) = 1:41,000 = 2.4 per 100,000
[See VanKesterin1996analysis for these calculations]

Note that Bakker1993 began a trend towards viewing reported numbers as absolute. This report cements that trend, in remarks in the introduction (p 590).

This report also makes a serious error in its reporting of the results of Tsoi88: It reports them as 1:9,000 males and 1:27,000 females (instead of 1:2,900 and 1:8,300)

[10]

Questionnaire sent to all GP MD’s in Scotland in April 1998.

Each asked to provide (i) number of clients and (ii) number of TS among them.

A total of 273 “gender dysphoric” patients were counted. Data broken down by mf and fm, and results tabulated for the various prevalences (TS, HT, SRS).

Report does not indicate awareness that the data only provides lower bounds. Presents the data as if absolute values.

Note that > 1/3 of TS clients had presented within the prior 12 months. This suggests that these clients were part of a ramp-up in INCIDENCE in help-seeking during the time when the UK medical system was beginning to more widely provide trans medical care under the gov’t health system.

[11]

By using initial data from the 1960’s and ‘70’s obtained from Harry Benjamin, M.D., and then tabulating counts of surgeries performed by top surgeons after that, [Conway01] presented estimates of the number of mf sex reassignment surgeries performed on U.S. residents during the preceding four decades, finding that the likely prevalence of P(SRSAmf) in the U.S. was at least 1:2,500 at the time.

From this result, [Conway01] hypothesized that the likely lower bound on the prevalence of inherent male-to-female (MtF) transsexualism in the United States was on the order of ~1:500 and might be even larger.

[12]

[Kelly01] applied the same methodology as in [conway01] to the UK, and found from government tabulations of SRS’s that the P(SRSAmf) in the UK was at least 1:3,750.

[13]

Sam Winter of the University of Hong Kong describes an interesting new method for triangulating on the prevalence of transsexualism in his paper entitled "Counting Kathoey" [Winter02].

[Winter02].estimated the prevalence of socially transitioned women, i.e., P(STAmf), in Thailand by counting “kathoey” among women passers-by at a number of public locations. Counts were done by “kathoey” who were top-experts at "reading kathoey".

[Winter02] found that approximately 6 per 1000 (i.e., 1:167) Thai women passers-by were mf social transitioners. I.e., P(STAmf) ≈ 1:167.

If even only a modest fraction (say 1/4) of these transitioners are transsexual, this would suggest a lower bound on inherent mf transsexualism on the order of 1:700 or so.

[14]

This report explores the trends in annual population incidence and mf/fm ratio of numbers of trans people seeking SRS and getting SRS from the official Swedish gender-reassignment program.

In recent years there has been a trend towards a very high percentage of those seeking SRS actually getting SRS in Sweden, when compared with the early years of the Walinder report.

This suggests that few people approach the gov't program unless they are far along, and are ready to seek SRS via the country's medical system.

The report tabulates earlier reports' results as if commensurable with this report (even though they are not).

The report shows no awareness of the numbers being lower bounds, or of people seeking and getting treatment (hormones, SRS) outside of the Swedish program.

[15]

[DeCuypere06] conducted a survey of Belgian surgeons to obtain a count of individuals up to 2003 who had undergone SRS by Belgian surgeons since 1985, with the following results:

P(SRSmf) [DeCuypere06] = 1:12,900
P(SRSfm) [DeCuypere06] = 1:33,800

[DeCuypere06] refers to these prevalences of sex reassignment surgery as “the prevalence of transsexualism”, and therefore follows the pattern set by [Bakker88] and [VanKesteren96] of implicitly defining a “transsexual person” as being “someone who has undergone SRS”. The report does not indicate that these are lower bounds on the prevalence of SRS, and carries out the results to three significant figures.

From [DeCuypere2006] we note that 292 mf patients underwent SRS in Belgium over the 18 year period from 1985 through 2002. Thus there were on average 16.2 mf patients receiving SRS per year. The number of males born in Belgium during 1968 (i.e. 34 years before 2002) was 72,430, and the average age of the SRS’s was ~ 34 years, taking into account a 2 to 2.5 year delay between first consultations and SRS [DeCuypere2006].

By applying Equation 4 in [Olyslager and Conway 07], the inherent prevalence of SRSmf is estimated from this data as follows:

P(SRSImf) derived from [DeCuypere2006] = 16.2/72,430 ≈ 1 in 4,500

[16]

In [Winter06], Winter conducted a survey of 195 Thai transgender females (mf), and compiled a very detailed demographic profile of this sample - tabulating the percentage of participants reporting various transition events (such as living in female clothes (ST), taking hormones (HT), undergoing SRS, etc.) and the ages of onset of those actions.

By combining the results of [Winter02] and [Winter06] and using Eqn. 5 of [Olyslager and Conway 07], we can estimate the prevalence of active and inherent SRS in Thailand as follows:

P(SRSAmf) = (0.277) (1:167) ≈ 1:600
P(SRSImf) = (1:600) / (73-24.1/73) ≈ 1:400

Winter also determined that 48.2% of those who hadn’t had SRS would like to have SRS. Therefore, 62.5% of the overall sample were transsexual under Walinder’s definition:

I.e., the % having had or desiring SRS = 0.277 + (0.482)(1.00 – 0.277) = 0.625 = 62.5%

By combining the results of [Winter02] and [Winter06], and taking 18.4 as an average of ST (taking that as “active transsexualism” in the instances of transsexualism), we can estimates the prevalence of active and inherent transsexualism in Thailand as follows (noting that these are first-order estimates, and may be influenced by a higher prevalence of kathoey in Thai cities than in the overall country):

P(TSAmf) derived from [Winter02, Winter06] = (0.625)(1:167) ≈ 1:270
P(TSImf) derived from [Winter02, Winter06] = (1:270) (73-18.4/73) ≈ 1:200

[17]

[18]

Incidence per population of mf and fm name changes in Sweden circa 1970.
Some unknown fraction will have had SRS.

Ip = 0.2 per 100,000 over 15 years of age for both mf and fm

[19]

[20]

[21]

[22]

This paper uses the same criteria as Waliner for identifying transsexual people.

However, it actually counted records of SRS's from the Dept of Obst and Gyn at the National Univ., plus a few reports from two private surgeons.

Thus this paper is reporting on the active Prevalence of SRS, rather than the prevalence of transexualism.

Tsoi88 reports them occurring at an average age of 24.1 (in a population having a life expectancy of ~ 80). By using Equation 5b in [Olylager and Conway 07] we can infer a likely lower bound on P(SRSImf), i.e., the prevalence of those who’ve had mf SRS plus those tracking towards SRS, as follows :

P(SRSImf) derived from [Tsoi88] ≈ 1:2,900 (80 / (80 – 24.1) = 1:2,000

[23]

[24]

[25]

[26]

[27]

[28]

[Kelly01] applied the same methodology as in [conway01] to the UK, and found from government tabulations of SRS’s that the P(SRSAmf) in the UK was at least 1:3,750.

[29]

[30]

[31]

[32]

[33]

Follow-on to Walinder68, with data on patterns of residence change.
No new info re Prevalence.

[34]

This brief paper reports that of 716 diagnoses among patients "requiring sexological diagnoses" in Cracow between '74 and '78, 0.3% of the males were transsexual and 8% of the females were transsexual.

They advance this data as evidence that the ratio of mf to fm transsexuals is very low in Poland.

[35]

Comment re Godlewski's report - suggesting reasons re differences in apparent sex ratio of trans people in Poland vs western europe

[36]

Paper about "psychosocial" issues and theories.

Not relevant to prevalence analyses

[37]

Review article focussing on comparisons of past results - no new data or results re prevalence.

[38]

This is a lengthy review article on the overall topic of transsexualism.

It includes a discussion of trans prevalence and sex ratio on p 320.

It does not present new results on prevalence, except for referring to unpublished data on sex ratios by the first author.

[39]

This report provides data on the numbers and mean ages of mf and fm people seeking help from four specialized gender clinics in Germany - comparing the numbers in past years with more current years.

It discusses the trends in ratios of mf to fm, and questions re these ratios.

It does not provide overall estimates of prevalence.

[40]

This is a short report that summarizes the prior work in prevalence.

In this report that the author tabulates all prior work as if it were normalized to same definitions and same measureables (i.e., commensurate), rather than needing normalization prior to comparisons.

The author carried over this incommensurate form of tabulation in the later report dcuypere2006.

[41]

Overview of some gay/trans issues. Summarizes past work in trans prevalence.

No new data, except for inclusion of unpublished Belgian work of Carael, 2004 that lie in same range as prior reports.