Using Prescriptions as a Proxy of Disease: An Indicator, Not an Indication

Great care must be taken when using prescription data as an indicator of disease.  When diagnosis information is not available prescription data can only be used as an indicator, not an indication.  Accurate definitions of the incidence and prevalence of some diseases in the population are lacking.  For example, understanding the burden of depression and anxiety is notoriously complex, mainly due to the difficulty surrounding the definition of these disorders and in identifying the sufferers.  Depression and anxiety are thought to affect an estimated one in four people in their lifetime with depressive disorders ranked as the third leading contributor to the global burden of disease.¹ But there is no population register of depression sufferers and the stigma surrounding mental health means many individuals will not admit to having a disorder.^2,3 So how can we study those individual suffering from depression or anxiety?

Surveys can provide a reasonably cost-effective way of gaining insight into the health of the population but they are labour intensive and subject to a variety of biases, the most important of which is responder bias.  Individuals who respond to surveys differ greatly from those who do not respond.  In terms of those characteristics that influence survey participation, evidence suggests that females,^4,5 older respondents,^4,5 those from higher socioeconomic backgrounds,⁶ those who are employed,^4,6 and those who are married are more likely to consent.⁶  Many of these characteristics are also associated with better mental health.^7-9  A study by Vercambre & Gilbert (2012) found that persons with mental complaints are less likely to respond to surveys, especially if the survey focus is on mental disorder.¹⁰

The rating scales of mental ill health within surveys are also subject to a disease threshold value, an arbitrary cut off differentiating a “normal” score from an “abnormal” score.  It is difficult to determine accurate incidence and prevalence rates because the case score identified in one study may not correspond to that used in another.  Even when studies use clinical interviews, what one practitioner diagnoses as depression another may not.

Prevalence estimates based on treated populations, such as psychiatric admissions, often underestimate the true population prevalence as not all individuals with a mental health disorder will be hospitalised.

However, most common mood disorders are treated pharmacologically with antidepressant, anxiolytic and/or hypnotic medication and many researchers are now turning to administrative prescribing data sources for information on population mental health.^11-14  Prescription data offers a readily available, affordable, quantifiable, population wide[i] measure of drug utilisation.  The problem with prescribing databases is that most do not contain information on the clinical reason or indication for prescribing.

Does prescription equal illness?

Unfortunately specific drugs do not always equate with specific illnesses. For some prescription medications this is the case.  For example, insulin is prescribed for diabetes and there is a consistent one to one mapping of prescription and disease.  Only individuals with diabetes will receive a prescription for insulin.  The same can be said for anti-obesity medication.  Individuals only receive anti-obesity medication when they are obese, so there is a one to one mapping of disease and disorder.  When considering medications that are indicated for the treatment of more than one disease, the use of prescribing data gets complicated.  The Grief Study will use prescription of an antidepressant, anxiolytic or hypnotic as an indication of mental ill health.  But what implications will this have on our findings?

The mentally well on drugs for mental ill health

Antidepressants, anxiolytics and hypnotics are now used for a wide variety of diseases including chronic pain, fibromyalgia, chronic fatigue, migraine, irritable bowel syndrome, insomnia and eating disorders.^15,16 Some would argue all of these disorders are somatisations of mental disorders, but this cannot be assumed for all cases.  GPs feel antidepressants, especially SSRIs, are safe drugs that can be used for many indications and it is believed over-treatment to healthy individuals will have less side effects than under-treatment of those in need.¹⁷ Occasionally an individual can receive a “one-off” anti-anxiety prescription for a stressful event or situation such as a job interview or long-haul flight.  This does not mean they suffer from a mental disorder.

The mentally unwell not on drugs for mental ill health

In addition, not all individuals with a mental health problem will be treated pharmacologically.  Individuals who avail of alternative psychotherapy treatments and individuals who receive no treatment at all will be missing from prescription databases.

So when does prescription equal illness? A study by Gardarsdottir et al. (2007) linking prescribing data to primary care data analysed the reasons why people are prescribed antidepressants and developed an algorithm to determine disorder from prescription databases.¹⁵  The majority of individuals receiving an antidepressant have a diagnosis of depression.  Those aged less than 60 years were more likely to have a diagnosis of depression alongside a prescription for antidepressants when compared to those aged over 60 years.  Those aged over 60 years were more likely to be prescribed antidepressants for other conditions such as neuropathic pain.  Individuals with more than one prescription for an antidepressant over the study period were also more likely to have an indication of depression in their file.¹⁵ Type of antidepressant prescribed is also a potential indicator of depression diagnosis.  The foremost indication for Selective Serotonin Reuptake Inhibitors (SSRIs) is depression.  Gardarsdottir et al. (2007) found that 73%of those individuals on an SSRI had a diagnosis of depression or anxiety whereas another study by Henriksson et al. (2003) found that 82% of those prescribed a SSRI had a diagnosis of depression compared to just 23% of those on tricyclic antidepressant (TCA).  The older drugs are more likely to be used for pain and anxiety disorders (Henriksson et al. 2003; Gardarsdottir et al. 2007).^15,18  The more information we have on the individual and prescribing history the more able we will be to identify those most likely to be receiving a drug for possible mental disorder.

Collating prescription data on all individuals who receive antidepressants, anxiolytics and/or hypnotics will inherently include some individuals who are receiving these medications for indications other than mental ill health and exclude some individuals who are suffering from mental ill health but are not receiving pharmacological treatment.  Misclassification bias will affect such a small proportion of individuals in a large population wide prescribing dataset that the only effect it is likely to have on the results is to underestimate the true magnitude of mental ill health in Northern Ireland.

When it comes to identifying a useful tool for the recognition of possible mental disorder in a population-wide cohort, prescribing data is the frontrunner. It’s not perfect. It’s a tool to aid our understanding.  But it’s the best tool we have and an accurate indicator of possible mental ill health.

References

1. World Health Organisation. Investing in Mental Health. Department of Mental Health and Substance Dependence, Noncommunicable Diseases and Mental Health, World Health Organization, Geneva. 2003
2. Cochran SV & Rabinowitz FE, Men and depression: Clinical and empirical perspectives. 2000. San Diego, CA: Academic Press
3. Addis ME & Mahalik JR. Men, masculinity and the contexts of help-seeking. American Psychologist. 2003;58:5–14
4. Eagan TM, Eide GE, Gulsvik A & Bakke PS. Nonresponse in a community cohort study: Predictors and consequences for exposure-disease associations. J Clin Epidemiology. 2002;55:775-81
5. Dunn KM, Jordan K, Lacey RJ, Shapley M & Jinks C. Patterns of consent in epidemiologic research: Evidence from over 25,000 responders. Am J Epidemiology. 2004;159(11):1087-94
6. Shahar E, Folsom AR & Jackson R. The effect of nonresponse on prevalence estimates for a referent population: Insights from a population-based cohort study. Atherosclerosis Risk in Communities (ARIC) Study Investigators. Annals of Epidemiology. 1996;6:498-506
7. Middleton N, Gunnell D, Whitley E, Dorling D & Frankel S. Secular trends in antidepressant prescribing in the UK, 1975-1998. J Public Health Med. 2001; 23(4): 262-7
8. Jenkins R, Lewis G, Bebbington P, Brugha T, Farrell M, Gill B et al. The National Psychiatric Morbidity Surveys of Great Britain – initial findings from the Household Survey.  Int Rev Psychiatry. 2003; 15: 29-42
9. Schoenborn CA. Marital Status and Health: United States, 1999-2002. Advance Data from Vital and Health Statistics. No.351. Hyattsville, Maryland: National Center for Health Statistics. 2004
10. Vercambre M & Gilbert F. Respondents in an epidemiologic survey had fewer psychotropic prescriptions than nonrespondents: an insight into health-related selection bias using routine health insurance data. Journal of Clinical Epidemiology. 2012; doi:10.1016/j.jclinepi.2012.05.002
11. NICE CG90: Depression. The treatment and management of depression in adults. The National Institute for Health and Clinical Excellence, NICE clinical guideline 90. 2009
12. Gardarsdottir H, Egberts A, van Dijk L, Sturkenboom M & Heerdink R. An algorithm to identify antidepressant users with a diagnosis of depression from prescription data. Pharmacoepi & Drug Safety. 2009;18:7-15
13. Pratt L, Brody DJ, Gu Q. Antidepressant Use in Persons Aged 12 and Over: United States, 2005-2008. NCHS Data Brief. No 76. October 2011
14. MacDonald TM, McMahon AD, Reid IC, Fenton GW & McDevitt DG. Antidepressant drug use in primary care: a record linkage study in Tayside, Scotland. BMJ. 1996;313:860-1
15. Gardarsdottir H, Heerdink R, van Dijk L & Egberts A. Indications for antidepressant drug prescribing in general practice in the Netherlands. Journal of Affective Disorders. 2007;98:109-15
16. Spettell CM, Wall TC, Allison J, Calhoun J, Kobylinski R, Fargason R & Kiefe CI. Identifying physician-recognised depression from administrative data: consequences for quality measurement. Health Serv Res. 2003;38(4): 1081–102
17. Mercier A, Auger-Aubin I, Lebeau JP, Van Royen P & Peremans L. Understanding the prescription of antidepressants: a qualitative study among French GPs. BMC Family Practice. 2011;12:99
18. Henriksson S, Boëthius G, Hakansson J, & Isacsson G. indications for and outcome of antidepressant medication in a general population: a prescription database and medical record study in Jämtland county, Sweden, 1995. Acta Psychiatr Scand 2003;108:427-31