Screening male prisoners for depression and anxiety with the PHQ-9 and GAD-7 at NHS Healthcheck: patterns of symptoms and caseness threshold

Higher levels of mental ill health have been found in prison populations compared to age, gender and education matched community samples with more than three-fold increases in all forms of depression and anxiety [1‐3]. The high prevalence of depression and anxiety disorders in prisoners may be related to the characteristics of prisoners such as poorer cardiovascular and other physical health, alcohol and drug misuse, guilt, poor family support and experience of trauma and violence in childhood and adulthood [4‐6]. The prison environment may also play a role in increasing depression and anxiety including lack of personal space, privacy, poor interpersonal relationships, social support, occupational roles and opportunities for exercise, occurrence of violence and coercion between prisoners, and lack of mental health access [3, 7].

The World Health Organisation [7] and The Bradley Report (2009) [8] recommended screening for mental disorder in the prison population to identify those who are a priority for clinical assessment by mental health professionals. Reasons for this include rising rates of suicide and self-harm [9], recidivism associated with poor mental health [10], and insufficient staff trained in mental health to meet the demand among prisoners [7, 10]. In England prisoners could be screened for depression and anxiety disorders alongside screening for risk of cardiovascular disease through NHS Healthchecks [11] to improve both mental and physical health together. However, self-rated screening instruments that are commonly used for the most prevalent mental disorders in prisoners and the community, namely depression and anxiety disorders, such as the Patient Health Questionnaire (PHQ-9) [12] for depression and the Generalised Anxiety Disorder Assessment (GAD-7) [13] might be susceptible to false positive results [14‐16]. A meta-analysis of depression screening studies in prisoners estimated the prevalence of depressive episodes to be 19.1% (95% CI 15.0–23.9) but screening by self-rated measures of depression estimated depression to be present in 54.0% (95% CI 43.8–63.9%) [14]. There was considerable variation in the methods used to estimate depression by interview or self-rated measure, and the population varied by country, nature of prison, sentence or remand status, and selection by clinical or demographic factors. Previous research raises the possibility that symptoms commonly regarded as important for the diagnosis of depression might be commonly present among prisoners in the absence of depression such as difficulties with sleep in a noisy environment, unfamiliar food, guilt over crimes and low self-worth [3]. On the other hand there might also be underreporting of suicidal ideation so that prisoners avoid close observation and further restriction on their activities that is often the response of staff to reported suicidal ideation [17].

Therefore the validity of using community cut-off scores on the PHQ-9 and GAD-7 for possible caseness of moderate and severe depression and anxiety disorders of 10 or more and 15 or more merit further examination. Possible methods might include assessment for the presence of core symptoms of major depressive episode and generalised anxiety disorder are captured in prisoners scoring above caseness, whether there are expected associations with demographic and clinical factors such as alcohol use and cardiovascular health, and whether clusters of symptoms emerge in cases that are consistent or inconsistent with diagnostic criteria for major depressive episode and generalised anxiety disorder. There are also alternatives to detecting cases of depression and anxiety that can be utilised in relation to the PHQ-9 and GAD-7 such as varying cut-off scores, use of the PHQ-9 diagnostic algorithm [12] and examination of symptom profiles through cluster analysis.

A cross-sectional prevalence survey was conducted in prison healthcare services within the East Midlands [18]. In the period of data collection from September 2017 to January 2019, there were 13 male prisons in the region, with the healthcare services at six prisons agreeing to contribute to the research. These prisons varied in prison category, each housing prisoners of different sentence length.

The study drew on data collected at the time of the NHS Healthcheck [18]. All prisoners (regardless of sentence length or time served) who were deemed eligible for the NHS Healthcheck in Prisons were invited to participate (aged between 35 and 74 years old and with no exclusion diagnosis). Eligibility was sought using clinic reports from SystmOne, an NHS electronic medical record system, where those ineligible were subsequently filtered according to NHS Healthcheck eligibility criteria; those excluded were prisoners with established cardiovascular disease: coronary heart disease; strokes; transient ischaemic attacks; diabetes; atrial fibrillation; heart failure; peripheral artery disease and chronic kidney disease and those prisoners already taking statins. Each prison had a new report run every 2–3 weeks to allow for new prisoner receptions and to discount release or transferred prisoners from being invited to clinic.

At the clinic appointment, physical measures were collected from prisoners attending the NHS Healthcheck which included age, ethnicity, height, weight, body mass index (BMI), waist circumference, systolic and diastolic blood pressure (BP), smoking status, family history, alcohol intake using the Alcohol Use Disorders Identification Test (AUDIT) [19]. Physical activity was recorded using the General Practice Physical Activity Questionnaire (GPPAQ) [20]. Blood tests for creatinine, plasma glucose, lipids and HbA1c were requested. Age, ethnicity and last known home address were recorded. Index Multiple Deprivations (IMD) code was applied to the postcode of the prisoner’s last known home address; IMD 1 being the most deprived and IMD 10 the least [21]. A large percentage of prisoners were of no fixed abode (NFA) prior to incarceration and could not be assigned an IMD code. Sentence length was recorded where available.

We found a high prevalence of prisoners scoring above conventional thresholds for moderate and severe depression and anxiety on the PHQ-9 and GAD-7. However, around half of prisoners with moderate depression on the PHQ-9 did not exhibit core symptoms of depression on a daily basis, and a quarter of prisoners with moderate anxiety on the GAD-7 did not have daily core symptoms of worry. Ninety-six per cent of prisoners with severe anxiety on the GAD-7 had daily core symptoms of anxiety suggesting that a GAD-7 score > 15 has high specificity for the presence of generalised anxiety disorder. However, one quarter of prisoners with severe depression on the PHQ-9 did not have daily core symptoms of depression. Increasing threshold scores of caseness on screening measures for anxiety and depression [27] from > 10 to > 15 on the GAD-7 would provide a high specificity for generalised anxiety disorder but not on the PHQ-9 for depression. Given the possibility of under-reporting as well as over reporting of some symptoms such as suicidality, a caseness threshold of 10 or more may be satisfactory for the GAD-7 but not for the PHQ-9. The diagnostic algorithm suggested that only 17.9% prisoners had a depressive episode which is in the expected range for caseness for a depressive episode (15.0–23.9%) when standardised psychiatric interviews were used in previous studies [14]. However, the diagnostic algorithm method tends to have low sensitivity [28] for depressive episodes so it may miss some cases, including some who may be at risk of suicide and self-harm. Therefore consideration might be given to other methods of utilising the PHQ-9 and GAD-7 since both the PHQ-9 and GAD-7 showed excellent internal consistency [23] in this sample of prisoners and their completion was highly acceptable to prisoners. The problems of case detection on the PHQ-9 using the community caseness threshold of 10 is not due to the low internal reliability of these measures among prisoners.

We found many univariate associations between no/mild and moderate case levels of depression and anxiety and younger age, shorter and longer sentence length, prison, higher alcohol consumption and lower cardiovascular risk. Many of these cardiovascular risk and prison or sentence factors were strongly related to age. Depression and anxiety caseness were not independent of each other either with 27.1% of the sample scoring 10 or more on both the PHQ-9 and GAD-7. As a result in the multinominal logistic regression analysis only prison and alcohol consumption were associated with caseness thresholds of > 10 and > 15 on the PHQ-9 and additionally young age was associated with the caseness threshold of > 15 on the PHQ-9. The relationship between alcohol consumption and PHQ-9 score was complex. High alcohol consumption was associated with severe depression but also with no or mild depression compared to moderate depression. Alcohol lowers mood and not being exposed to it might lift people out of milder states of depression but not out of severe depression [29]. No variables were associated with caseness thresholds of > 10 or > 15 on the GAD-7. However, the relationship of anxiety symptoms to demographic factors, alcohol use, cardiovascular disorders, sentencing and prisoners have been much less explored than for depression in prisoners or other samples. Therefore, the lack of an association between these variables and the severity of anxiety in this sample is open to interpretation. The lack of an association between the severity of depression and factors such as social deprivation and higher cardiovascular risk was not expected, possibly because many prisoners were of no fixed abode so not contributing to assessment of social deprivation, and the sample was predominantly white British in ethnicity [18].

One of the two remand prisons and a near parole prison for prisoners with only 3–6 months left to serve of a longer sentence showed increased rates of caseness on the PHQ-9 compared to the other prisons, including another remand prison. These prisons and the other remand prison with lower rates of caseness on the PHQ-9 had high ‘churn’ (turnover) rates, with prisoners moving through these establishments at high rates. Precision in relation to the denominator mean that estimates of prevalence of caseness is less certain in prisons with high churn, but the churn also means less opportunity to establish social support which has been related to common mental disorder in prisoners [3]. Both remand and sentenced prisoners report high rates of victimisation in prison and low rates of social support available to them in prison and the community [3]. The remand prison with high PHQ-9 caseness was a large prison associated with relative overcrowding compared to the other remand prison with lower caseness of depression. There were also high rates of caseness on the PHQ-9 and GAD-7 among a small sample of only 25 prisoners with life sentences but this was too small a sample to draw firm conclusions.

Given previous concerns about false positives when screening prisoners for mental health problems with community derived thresholds [14‐16], we conducted a dendrogram of scores of 3 or more on the core items of the PHQ-9 and GAD-7 in all those who scored > 10 on the PHQ-9. We found five clusters, the largest of which (cluster 3) had daily core symptoms of depression with suicide ideation, poor concentration, feeling slowed down or restless, irritability or fear of something bad happening. This cluster might consist of prisoners with symptoms suggestive of a depressive disorder meriting further clinical examination. In relation to the other clusters, cluster 1 contained prisoners scored highly on the first four worry and related symptoms of GAD-7, including the three core symptoms, but not on the remaining GAD-7 items or PHQ-9 items. A previous study indicated that over 40% of male sentenced and nearly 60% of remand prisoners reported frequent worry [3]. Cluster 2 related to Item 5 (altered appetite) reflecting a prison diet that is rather different from their previous diet. Cluster 4 was an outlier consisting of worthlessness and guilt (item 6 on the PHQ-9) which was frequently scored highly (25.1%). Guilt may have been associated with the crime committed or letting down family while low self-esteem may have been prominent throughout their lives, not necessarily related to current feelings [30, 31]. Cluster 5 consisted of high scores on poor sleep and fatigue; insomnia and fatigue are frequently reported by prisoners in noisy environments [3].. Thus there are possible environmental, guilt and stress factors in a prison setting influencing the scoring of a number of items on both the PHQ-9 (e.g. items 3–6) and GAD-7 (items 1–4), potentially inflating PHQ-9 scores, and GAD-7 scores to a lesser extent, in a prison setting.

We have confirmed previous suspicions that using conventional community cut-off scores on self-rated measures of depression tends to overdiagnose depression. Using a threshold of 10 or more on the PHQ-9 to detect depression caseness seems to overinflate detection rates because it captures too many prisoners struggling with distress from poor sleep and fatigue, poor appetite, guilt or low self-worth who do not have core symptoms of a depressive episode nearly every day over the previous 2 weeks as defined by DSM-V [22]. Although these problems have been documented as distressing to prisoners before [3], this study is the first to provide evidence that such problems may be contributing to an overdiagnsosis of depressive episodes using a community threshold of caseness for depression of 10 or more on the PHQ-9. A positive score on the PHQ-9 diagnostic algorithm, a cut-off score of 15 or more on the PHQ-9, or a symptom profile similar to cluster 3 on the PHQ-9 and GAD-7 in prisoners scoring 10 or more on the PHQ-9 are worth exploring singly and in combination as a method of efficiently finding prisoners who may have a depressive episodes in further research using a gold standard psychiatric interview. However, it is worth noting that one study which used multiple self-rated questionnaires only improved the specificity of detection of depression by a marginal amount [15].

Caseness for depression or anxiety disorder would result in a mental health assessment by a qualified mental health professional followed by an appropriate treatment plan. The latter might involve medication, psychological treatment or a change to social or environmental factors within the prison depending on the nature of the mental health problem and factors contributing to its maintenance.

PHQ-9 and GAD-7 scores may be inflated in prisoners by environmental, cognition and stress related factors. Further research using standardised diagnostic interviews should explore whether the use of higher threshold scores, diagnostic algorithm on the PHQ-9 and the pattern of items on the PHQ-9 and GAD-7 among prisoners scoring 10 or more among high scorers singly or in combination efficiently identifies cases of depressive episode and generalised anxiety disorder that merit further clinical examination by a mental health professional.