Key results
This is the first study describing the management of PTSD, anxiety and depression identified using a postal survey of survivors of critical illness. Nearly one quarter of patients who reported symptoms of anxiety, depression or PTSD following critical illness had a psychological diagnosis that pre-dated the first (3 month) questionnaire. As this was prior to the notifications being sent to GPs, the study itself could not have influenced diagnosis. However, in the three quarters without a pre-existing diagnosis a significant proportion (27%) received medical therapy or specialist referral. Without these notifications, some of these patients may have presented at the GP, but the notifications may have increased the timeliness of treatments.
Limitations
We have used two separate postal survey questionnaires. First, to assess the symptom burden amongst participants and then subsequently, as a means of assessing the outcome resulting from written notification of their GP. Survey questionnaires are subject to many biases and using two in series is likely to incur a double selection bias [
19]. Selection bias is well described amongst responders to postal questionnaires [
20,
21]. Following critical illness, it is also possible that the physical and mental barriers involved in responding to the study office may exert a degree of attrition bias from the outset. Response bias in the forms of acquiescence bias and extreme responding are also well-recognised features of postal questionnaires. However, from a pragmatic perspective, our question relates to the diagnostic yield of completed postal screening amongst those that do respond, rather than those that are lost to follow-up.
The determinants of response by GPs to our questionnaire are also subject to various forms of response bias. GPs were not incentivised to respond, differences in practice management, resourcing, record keeping and research involvement are all likely factors that will have influenced response. Despite these limitations, postal survey is a relatively unobtrusive, cost effective and efficient means of collecting data [
20]. The instruments used to identify the three causes of psychopathology are well validated.
We know from the ICON study and similar smaller studies that postal survey identifies the symptoms associated with psychological disease [
6,
14,
21]. The relativity low response rate amongst these studies means their results may not be representative of all survivors. It may also be that sufferers are over-represented in the responders (volunteer bias) [
14]. However, considering the relative ease and low cost associated with performing a postal survey, and the fact that three quarters of caseness was newly identified before the patient consulted their GP, it may allow early identification and intervention in many sufferers. In this sub-study, the proportions of patients with symptoms of PTSD versus anxiety/depression are higher than in previous ICON publications as the study design used a higher threshold applied to both HADS sub-scores to define severe symptoms than the lower caseness threshold used previously.
An overall response rate of 66% amongst GPs is as expected especially considering there was no obligation or incentive [
22]. That 37% of those who responded could not identify the original warning letter is certainly of concern and suggests the style or type of notification needs review. Failure of the postal services seems unlikely at this scale and did not occur in the ICON study when sending questionnaires to participants who had already responded. There are known to be large variations in the accuracy and completeness of the clinical information stored in electronic patient records [
23]. It is possible that the interface between paper notification letters and progressively more electronic general practice records contritubuted to failure to find the original letter. Regardless of the explanation for this result, this study highlights some of the challenges involved in establishing high-quality exchange of medical data between interested parties.
Both instruments used have been validated and their psychometric properties described. HADS has been used extensively in both survivors of critical illness and much wider patient populations. Alerting was based on the definition of `severe` symptoms (score ≥ 15). However, systematic reviews of the use of HADS have shown that using the much lower cut-off value of (≥ 8) is associated with a specificity and sensitivity of 0.8 [
24]. This may account for the relatively high psychopathological burden and subsequent interventions amongst those with anxiety and depression. It also suggests that there is a significant unstudied population of patients with anxiety and depression within the ICON dataset.
Conversely, PCL-C has been predominantly used outside of the post ICU setting. A correlation of 0.93 between the total PCL score and structured interview with the Clinician-Administered PTSD Scale (CAPS) has been demonstrated with a diagnostic efficiency of 0.9 versus the CAPS [
25]. A score of 45 or greater on the PCL-C has been recommended as a cut off point for high PTSD symptom load [
13]. However, this can vary between populations, for example a threshold of 50 in breast cancer patients resulted in a sensitivity of 0.60 and a specificity of 0.99 [
13]. Regardless, of the threshold chosen, any instrument will ultimately result in a dichotomous outcome at the point of deciding whether to act upon the result. If PCL-C were to be used as a screening tool further work would be required in order to evaluate its yield and potential benefit/harm as well as cost.
Interpretation
GP intervention following notification of severe symptoms reported in a HADS or PCL-C questionnaire depends markedly on prior knowledge of disease. Receipt of an alert in a patient with a pre-existing diagnosis was followed by only 6% receiving further intervention. In contrast, in those with no prior diagnosis, nearly half received further assessment, with 27% receiving additional therapy or mental health referral. Our findings suggest there is a significant unmet psychopathological need in survivors of critical illness that routine use of HADS or PCL-C questionnaires could help detect.