Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
In the recent article “Population-adjusted numbers, demographics and mental health among children and adolescents referred to the Norwegian National Center for Gender Incongruence over two decades” [1], the authors describe a “sharp” increase in the number of referrals to the Norwegian National Center for Gender Incongruence (NCGI) between the years 2000 and 2022. Additionally, the prevalence of self-harm, suicidal attempts and psychiatric diagnoses for the referred population are presented.
While reading the article, we discovered uncertainties around the ethics approval, several serious shortcomings in the analyses, presentation of data, and inferences, questioning the validity of the drawn conclusions.
Anzeige
For simplicity, we will focus on the presented numbers for assigned-female-at-birth (AFAB) referrals, and the following main weaknesses in the article:
1.
Uncertainty whether an ethics approval was obtained.
2.
Misrepresentation of distinctive data features.
3.
Failure to consider pandemic-induced effects.
4.
Lack of considerations of temporal dependencies (random effects) in the data.
5.
Poor justification and presentation of statistical analyses.
Ethics
Before we focus on the contents of the article, we want to highlight that the current presentation of the ethics approval leaves open whether an official ethics approval had been obtained. The article states: “This study was approved by the Data Protection Officer (DPO) at Oslo University Hospital (OUS)”. However, DPOs cannot grant ethical approval in Norway. Instead, regional ethics committees are in place to evaluate medical and health-related research. We hence urge for an immediate clarification of the issue.
Distinctive features of the data require consideration: dependence, delay and hidden variables
The authors seem to ignore the crucial difference between gender incongruence and other types of common referrals at hospitals, such as temporary illness or illnesses that are not strongly linked to social acceptance. While initiating with a good reasoning of population adjustment, the authors fail to consider that the selected data points are in essence not independent. The latter issue limits the validity of regression models without random effects. In other words, generations, which are exposed to different societal views and experiences [2], are mixed up in the analyses.
Moreover, the start of data records corresponds to the time NCGI started offering treatments, making it reasonable to expect a “burn-in” bias effect.
Anzeige
Further, increased acceptance induces a delayed effect. Therefore, the data span 2000–2022 with the corresponding burn-in effect is severely limited.
Additionally, contrasting the presented data to demographic trends of trans-population prevalence is crucial. For instance, a study conducted by Statistics Norway, [3] found that the transgender population in Norway aged \(>18\) lies around \(0.4\%\), peaking among the younger generations (\(0.7\%\)). Moreover, \(1.5\%\) of respondents claimed that their legal gender did not match their own experienced gender. Similar proportions can reasonably be expected in youth. Other studies around the world (e.g., [4] for Swedish population) place these percentages between 0.1 to 2.7% depending on the country and political situation. The peak in referrals in the article lies around \(0.04\%\), far below the lower bound of the mentioned population estimates. Following, increased healthcare seeking behaviours should be promoted as positive and not described as “dramatic” as in the article’s reference [11].
Finally, a global trend of increased utilisation of healthcare services has already been described 25 years ago [5]. Such trend, also more recently observed (e.g., USA [6]), suggests that the in-person frequency of healthcare-utilisation might not be stable in the general Norwegian population over the recent decades. Not correcting for this potential per-capita increase in healthcare-utilisation via all-cause-referrals limits the validity of the findings.
All these data features are ignored in the presented article and cited literature, possibly due to the hegemonic practice to consider gender incongruence as yet another “disorder” and hence applying classical, yet inappropriate statistical analyses. In that spirit, all cited articles try to explain referrals from isolated psychiatric and demographic variables, while the true phenomenon is much more complex to describe and interpret.
Pandemic-induced effects must be accounted for
A dip in referrals during the pandemic years 2020–2021 (article: Fig. 2) is explained as “reduced working capacity at the CAPOCs and the health system in general”. The authors fail to consider the “double effect”: The reduced capacity and social confinement during the pandemic led to a higher load of referrals, as these were simply put on hold during the pandemic. All data points from 2020, 2021 and 2022 are therefore distorted by the pandemic. The model fit to these data forecasts hence inflated numbers, or what the authors call a “sharp increase”. Opposingly, the data point from 2019 could already hint at a moderate stabilization. Moreover, the 2023 AFAB prevalence from NCGI’s annual reports of 24.77 (population-adjusted), is outside the \(95\%\)-credible interval of the model presented in Fig. 2 (see Fig. 1). This renders the observed data to be highly unlikely (\(5\%\)), or the model performance to be extremely poor, providing more evidence for a mistaken conclusion about a steep increase.
Fig. 1
Figure 2 from Nyquist et al. with their 95% C.I. and the new values for the year 2023. Mann–Kendall tests provide p values of 0.558 for AFAB and 0.335 for AMAB during the period 2016–2023, indicating a constant trend during this period
We suggest that a more appropriate procedure would be:
1.
A prudent analysis ignoring these three data points.
2.
Assuming that much of the volume from 2021 and 2022 belongs to 2020, and for instance, averaging out the three observations.
Temporal dependencies necessitate adjustments for generation membership
The previous section highlighted the complexity and heterogeneity of the data, including the influence of past societal views [7]. Any regression analysis on the number of referrals requires independent responses, which cannot be assumed in this type of data set. Since transgender identity is not suddenly appearing and then disappearing, analyses should be adjusted for age and thereby generation membership to obtain independent data. An additional benefit is the resulting independence of the results from current societal trends, which have been shown to influence referral statistics [7]. Two scenarios to exemplify:
Scenario 1: Three children are referred in 2015, aged 15, 15, and 15, and three children are referred in 2016, aged 16, 16 and 16. Thus, the number of referrals in 2015 and 2016 is identical (\(0\%\) increase).
Scenario 2: Two children are referred to NCGI in 2015, aged 15 and 15 and four children are referred in 2016, aged 16, 16, 16 and 16. Thus, the number of referrals in 2015 and 2016 is doubled (\(100\%\) increase).
Both scenarios involve the same children, with the same gender identities, namely six children born in 2000. The author’s model would lead to different conclusions for each of the scenarios. The difference is due to timing, strongly linked to the environment, making generation-adjustments a necessity for clear analyses.
Nevertheless, the challenge of insufficient historical data to determine the “true” number of transgender children remains. Historical data are important as various events can influence referral numbers at various scales, for example:
1.
The initiation of treatment in 2000.
2.
Exclusion, stigma, and inherent transphobia in society and healthcare services.
3.
Social events that influence trends in both positive and negative ways: The 2013 Norwegian act on the prohibition of discrimination based on sexual orientation, gender identity, and gender expression, the 2016 removal of the sterilization requirement for legal gender change in Norway, removal of the “transexualism” diagnosis in 2019 by the WHO, the pandemic in 2020, the 2022 Pride Shooting in Oslo, or proliferation of conservative governments around the world passing anti-transgender laws [7].
These considerations and the resulting statistical complexities should be kept in mind in any statistical analysis that is temporal and strongly linked to social events, in contrast to disease or injury.
Anzeige
Statistical analyses require justification, transparency, and multi-perspective interpretation
The article takes many statistical assumptions for granted that are not accounted for. We provide a non-exhaustive list of examples:
1.
Lack of model descriptions: Why are trend and number of referrals modelled separately? (Binomial and logistic, respectively). What is the mentioned “upper limit”? This choice impacts confidence bands and should not be arbitrary.
2.
Model fitting: Pearson’s \(\chi ^2\)-test is employed in the absence of plausible assumption checks, for instance, on the necessity of Yates’ correction. The presented mixed/nested data require the use of mixed/nested/hierarchical models with random effects (hierarchies), time-series, or adjustments to the data as explained above. The authors also incur in overdispersion, a widely known fallacy, by using a binomial model rather than, e.g. a beta-binomial model to capture the data’s large dispersion. This might explain why so many values are located outside the models’ confidence intervals (article Fig. 3).
3.
Covariates, regression estimates, residuals and explanatory power: Descriptions of the suitability of the utilized model are absent. Are model assumptions met? What are the estimated coefficients and their significance? And most importantly, what is the amount of variability explained by the model? We suspect it is low given the multiple observations outside the confidence intervals in Fig. 3, and the 2023 data point from the annual report.
4.
Incomplete interpretations: Causal inferences between health/psychiatric assessments and referrals are convoluted by multiple factors. For example, minority stress can enhance health disparities [8], and as detailed in the article: intake assessments “could lead to more children and adolescents receiving psychiatric diagnoses compared to children and adolescents in the general population”. Despite not providing evidence, the authors only entertain the causal links between psychiatric disorders, being trans and self-harm and suicidal attempts, instead of alternative, evidence-based scenarios, such as transphobia increasing suicidal ideation [9].
Final remarks and conclusion
The main shortcomings in the article are the interpretation of the data set, incorrect model assumptions, unreported model coefficients, the lack of models’ explanatory power, and uncorrected pandemic effects, impacting the whole conclusion.
There was never a “sharp” increase. There has been an increase, also called “structural break”, still below population prevalence, which should be promoted as positive behaviour and independent of (or not caused by) psychiatric conditions.
We would warmly like to call researchers for prudence in drawing rushed and even wrong conclusions, followed by presenting incomplete interpretations which promote a narrative of being transgender as a disease.
Nyquist CB et al (2024) Population-adjusted numbers, demographics and mental health among children and adolescents referred to the Norwegian National Center for Gender Incongruence over two decades. Eur Child Adolesc Psychiatry 1–11
2.
Bragg S et al (2020) More than boy, girl, male, female’: exploring young people’s views on gender diversity within and beyond school contexts. In: Trans youth in education. Routledge, pp 100–114
3.
Dalen HB, Berglund F, Lillegård M (2024) Kjønn, identitet og seksualitet i kvantitative undersøkelser. Statistisk Sentralbyrå
4.
Åhs JW et al (2018) Proportion of adults in the general population of Stockholm County who want gender-affirming medical treatment. PLoS One 13(10)
5.
Hensher M, Edwards N, Stokes R (1999) International trends in the provision and utilisation of hospital care. BMJ 319(7213):845–848CrossRefPubMedPubMedCentral
6.
Mortensen K et al (2021) Trends in health care utilization and expenditures in the United States across 5 decades: 1977–2017. Med Care 59(8):704–710CrossRefPubMedPubMedCentral
7.
Indremo M et al (2022) Association of media coverage on transgender health with referrals to child and adolescent gender identity clinics in Sweden. JAMA Netw Open 5(2):e2146531CrossRefPubMedPubMedCentral
8.
Hoy-Ellis CP (2023) Minority stress and mental health: a review of the literature. J Homosex 70(5):806–830CrossRefPubMed
9.
GR Bauer et al (2015) Intervenable factors associated with suicide risk in transgender persons: a respondent driven sampling study in Ontario, Canada. BMC Public Health 15:1–15
Mit zwei bis drei Tassen Kaffee am Tag ist das Demenzrisiko um fast ein Fünftel reduziert, auch die kognitive Leistung ist dann deutlich besser als bei Kaffee-Abstinenz. Das gilt jedoch nur für koffeinhaltigen Kaffee. Ob Koffein tatsächlich protektiv wirkt, lässt eine Langzeitstudie aber offen.
In Europa ist die Zahl der Menschen mit ADHS-Medikamenten seit 2010 deutlich angestiegen, vor allem unter Frauen und Erwachsenen. Ein solcher Trend ergibt sich auch für Deutschland, allerdings auf vergleichsweise niedrigem Niveau.
Nach einem operativen Eingriff ist die Schlafqualität oft massiv beeinträchtigt. In einer US-Studie waren bei Risikopatienten nicht nur die Gesamtdauer des Schlafs, sondern vor allem auch REM- und Tiefschlafphasen deutlich verkürzt.
