Minimal use of ICD social determinant or phenomenological codes in mental health care records

Abstract Background We have previously argued that psychiatric diagnosis, by focusing on pathology, minimises the role of psychosocial factors. Despite suggestions that traditional diagnosis is the only way to access treatment services, we have recommended standardised use of existing codes for possible social determinants and precise description of psychological phenomena. Aims This study examines the current use of social determinant and phenomenological codes in mental health care records. Methods Data provided by a local NHS Trust included 21,701 cases with a first contact date between 01 January 2015 and 01 January 2016, 4656 of whom received a primary diagnosis. Results Overall, codes for possible social determinants were used on only 43 occasions, for 39 individuals (0.8% of the 4656 people receiving a primary diagnosis). Comparison with relevant baseline frequencies revealed a highly significant under-reporting of key social determinants. 19 cases (0.4% of 4.656) used codes for precise description of specific psychological phenomena. Conclusions Available ICD codes for social determinants and specific psychological phenomena are under-used in clinical practice.


Introduction
There is overwhelming evidence for the role of social determinants -"the conditions in which people are born, grow, live, work and age" (WHO Commission on Social Determinants of Health and & World Health Organization, 2008, p. 1) in the development and maintenance of mental health problems. Poverty is a risk factor for nearly every mental health problem (Allen et al., 2014), as are adverse experiences in childhood (Mandelli et al., 2015), living alone (Bijl & Ravelli, 2000), and homelessness (Fazel et al., 2008).
ICD-10 (and its successor, ICD-11) includes many nondiagnostic codes recommended for the description of a large number of circumstances or problems which influence health (World Health Organization, 1993). The majority of these fall within "Z codes" or "factors influencing health status and contact with health services". These codes can further be subclassified as physical health threats (such as "genetic susceptibility to disease") and, relevant for the current discussion, "potential health hazards related to socioeconomic and psychosocial circumstances". These include codes related to "employment and unemployment" (e.g. "discord with boss and workmates"), "housing and economic circumstances" and the "social environment" (e.g. "homelessness" and "living alone") "negative life events in childhood" and "other psychosocial circumstances". Relevant additional codes fall within sections T66-T78, "other and unspecified effects of external causes" (including physical, sexual, and psychological abuse) and section X85-Y09, "assault" (including sexual assault, neglect and abandonment).
These codes represent an evidence-based list of known social determinants of mental health problems. Care would improve if these factors were better acknowledged (Borrell-Carri o et al., 2004), and record-keeping would match the biopsychosocial model as well as inform policy agendas (Allsopp & Kinderman, 2017;Braveman & Gottlieb, 2014). In 2012, a leading group of social psychiatrists (Bracken et al., 2012) argued that mental health care needed to reform, to take better account of social determinants. Similarly, the United Nations Special Rapporteur Dr Dainius P uras (P uras, 2017) argued that mental health problems are strongly linked to early childhood adversities, inequalities and abuse, and argued for a "revolution" in mental health care; a shift in focus from "treatment" towards a more fundamental social basis for care. These kinds of visions for care (see, for instance Kinderman, 2019) take, as their first step, the recognition and recording of these social determinants.
In addition to the use of codes for the social determinants of mental health problems, ICD-10 (as well as ICD-11 and the DSM franchise) also includes non-diagnostic phenomenological codes. Academics and clinicians (British Psychological Society, 2011;Kinderman & Allsopp, 2018) have suggested that specific distressing experiences or phenomena, as defined by individuals themselves, could form the basis of clinical records and care planning. In the ICD system, phenomenological descriptions fall under the subsection of "symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified". These include specific phenomena (such as low mood, anger, or hearing voices) which have specific ICD codes as well as appearing elsewhere as symptoms of diagnosed disorders such as "major depressive disorder" or "schizophrenia". Importantly for this study, therefore, both ICD-10 and ICD-11 offer the possibility of codes such as; auditory hallucinations (R44.0), anhedonia (R45.84), suicidal ideations (R45.851), or emotional lability (R45.86).
Phenomenological codes could avoid the lack of reliability, validity and utility of traditional diagnoses, and could help clarify epidemiological links between social determinants and mental health outcomes (Stevens et al., 2004). It is worth reiterating that the use of such codes (both social determinants and phenomenological) are explicitly recommended in both the ICD and DSM systems, for instance in the call to record a; "concise summary of the social, psychological, and biological factors that may have contributed to developing a given mental disorder" (American Psychiatric Association, 2013, p. 19).
However, in practice, it appears that both record types are not well represented in the academic and clinical literature. Venzon et al. (2019) reported that social and demographic data were rarely found in electronic health records. When these data were captured, they were often manually entered and related to local community and neighbourhood characteristics. Similarly, Torres et al. (2017) found that information related to social determinants (and more specifically post-discharge risk factors) had been recorded in only 1.6% of 35,600,000 hospital discharges. Therefore, while both the ICD and DSM systems include the possibility of systematic recording of data on social determinants, these opportunities are rarely used in clinical practice.

Aims
The current study was designed to assess the extent to which the available social determinant and phenomenological codes were used in clinical practice. Because of the importance of these known social determinants of mental health problems, the frequency of their use was compared to estimates of population prevalence. Furthermore, metaanalyses have demonstrated clear relationships between social determinants such as trauma, intimate partner violence, or adverse childhood experiences and specific mental health problems; e.g. depression (Mandelli et al., 2015) and anxiety (Lindert et al., 2014), where childhood trauma is over twice as common as in the general population, and intimate partner violence is a major risk factor for postnatal depression (Wu et al., 2012); obsessive compulsive problems (Miller & Brock, 2017), where the severity of problems is associated with the degree of trauma experienced in childhood; non-suicidal self-harm (Liu et al., 2016) , where people surviving trauma in childhood are significantly more likely to injure themselves in adulthood; and even psychosis (Varese et al., 2012), mania or hypomania (Palmier-Claus et al., 2016), where, again, childhood trauma predicts later mental health problems. Therefore, the frequency of use of these relevant codes in relation to specific diagnoses was also examined.

Method
The current study compared the prevalence of particular social determinants of mental health problems in the general population and those seeking help for mental health problems with the recorded prevalence in a UK NHS Trust dataset. In addition, the frequency of use of specific phenomenological codes was analysed. UK NHS Trusts use a variety of electronic patient record systems to record both mental and physical health information for their service users. The Trust in question used the ePEX system, an electronic healthcare records system designed both for internal Trust clinical activity and for mandatory reporting to commissioners and regulatory bodies. A request was made for any and all ICD codes used for any person included on the system between the specified dates, which were then analysed by the authors.

Sample
The population of interest was all cases of the NHS Trust responsible for mental health care in a city region in the North West of the United Kingdom, with a first contact date between 01 January 2015 and 01 January 2016. The NHS Trust responsible for most of the city's mental health care is a specialist mental health NHS Trust, with (in 2019) over 7600 staff, based in 85 sites, serving a population of almost 11 million people. In 2015, the period for which this data was available, the Trust had 3448 full-time equivalent staff. The Trust provides specialist inpatient (including high secure) and community mental and physical health services; these include services for older adults and people living with dementia, learning disabilities, addiction management and acquired brain injury services. The catchment area is a diverse city like many others, however the rates of people living in poverty and other markers of social deprivation are amongst the highest in the UK (Parkinson et al., 2016).
All 21,701 cases were fully anonymised, with no information about names, age, ethnicity or gender. Only 4656 cases included diagnostic information or records of social context. The remaining 17,045 cases included only data on the dates of episodes of care.
All records referred to individual people who had received some form of clinical service from the Trust. Duration of care episode (calculated from first and last contact dates) was unavailable for 1663 individuals, because they were still active patients of the Trust. For those whose cases had been closed, 1882 had a first and last contact date on the same day, most likely indicating a consultation that did not lead to ongoing clinical activity. Of the remaining 18,156 individuals, the care episode lasted for a mean of 253.8 days (SD ¼ 261.9).

Analysis
First, all 21,701 cases on the database were tabulated by whether the individual had or had not received any of the 357 ICD-10 mental and behavioural disorder ("F-code") diagnoses, recorded at any point after their first contact date. Each individual may have received either none, one or more than one of these diagnoses. The cases were then also tabulated by whether they had received or not received ICD codes related to social determinants. This analysis included each of the 120 factors influencing health status and contact with health services (for brevity referred to as "Z-codes"), again recorded at any point after their first contact date. Individuals may have been given either none, one, or more than one, of these codes.
The ICD system is flexible and heterogeneous (Allsopp et al., 2019). Therefore, ICD-10 codes referring to psychosocial factors (e.g. Z61.4 "problems related to alleged sexual abuse of child by person within primary support group"), are supplemented by codes for physical health issues (e.g. Z86.7 "personal history of diseases of the circulatory system") and service access (e.g. Z91.1: personal history of noncompliance with medical treatment and regimen). Further complicating matters, diagnostic and quasi-diagnostic codes included those categorised by ICD-10 as "mental, behavioural and neurodevelopmental disorders" ("F-codes"), but also "diseases of the nervous system" ("G-codes"), "congenital malformations, deformations and chromosomal abnormalities" ("Q-codes"), and "symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified" ("R-codes"). Finally, as an added complication, all these ICD codes, including both social determinants and phenomenological codes (i.e. "factors influencing health status and contact with health services" and "symptoms and signs not elsewhere classified") also occasionally appeared as primary diagnoses (in the terminology of the Trust).
In this analysis, therefore, all relevant ICD codes were included, and have been referred to both in text and, comprehensively, in the tables. This may appear overinclusive, but allows readers both to see the complexity of the system and draw their own conclusions as to the utility of the encoding and the conclusions drawn.

Primary diagnosis
A total of 4656 cases (21.5%) included a primary diagnosis. These can be seen in Tables 1-5. In this database, 769 people (3.5% of the total sample, 16.5% of those with a primary diagnosis) received a diagnosis of an "organic, including Alzheimer's disease, unspecified 146 0.7 G31.0 Circumscribed brain atrophy 7 0 G31.8 Other specified degenerative disorders of nervous system in diseases classified elsewhere 25 0.1 symptomatic, mental disorder' (e.g. F01.9 'vascular dementia, unspecified"). A primary diagnosis of a "disease of the nervous system" (e.g. G30.1 "Alzheimer's disease with late onset") was made for 1550 individuals. A primary diagnosis of a "mental and behavioural disorder due to psychoactive substance use" (e.g. F10.0 "acute alcohol intoxication") was made for 764 individuals. A diagnosis of a "mental, behavioural and neurodevelopmental disorder" (conventional mental health diagnoses such as "paranoid schizophrenia" F20.0) was made for 2221 individuals (10.2% of the total sample, 47.7% of those with a primary diagnosis), and a diagnosis of a learning disability (termed "mental retardation" in ICD-10) or similar diagnosis (such as; G93.3 or Q90.9 "Down"s syndrome') was made for 96 people. Five people were diagnosed with "disturbance of activity and attention" (F90.0), and one with "postviral fatigue syndrome" (G32.8).

Mental health phenomena
Also included as "primary diagnoses" in this dataset were several ICD codes which are not conventional bio-medical diagnoses. These generally fell under the ICD heading of "symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified". Thus, two people were assigned the quasi-diagnostic code for "auditory hallucinations" (R44.0) and 14 people were assigned the quasi-diagnostic code for "other symptoms and signs involving emotional state" (R45.8). One person was assigned the quasi-diagnostic code for "other symptoms and signs involving appearance and behaviour" (R46.8) and one the quasidiagnosis of "strange and inexplicable behaviour" (R46.2). Finally, one person received the quasi-diagnosis of "stress, not elsewhere classified" (Z73.3). In total, 19 people (0.1% of 21,701 or 0.4% of 4.656) received these quasi-diagnostic codes as (or in place of) primary diagnoses.
Closer analysis of the use of non-diagnostic codes depended largely on manual examination, because of the heterogeneous nature of the ICD system. For this analysis, therefore, we manually examined all entries which included any ICD code. In this database, ICD "R-codes" ("symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified") were used on a total of 66 occasions, and "I-codes" ("diseases of the circulatory system") were used on 363 occasions.

Social determinants
Overall, "Z-codes" were used on 1507 occasions, for 892 different individuals (4.1% of the whole dataset of 21,701 individuals, or 19.2% of the 4656 records with diagnostic information). Table 6-8 summarises these data. As can be seen, these Z-codes fell into three broad categories; physical health issues (n ¼ 228), risk-related factors (n ¼ 1180) and social determinants (n ¼ 43, recorded for 39 individuals), the principal focus of this paper. Physical health factors included such issues as; Z85.4 "personal history of malignant neoplasm of genital organs" (recorded for four individuals). The risk-related factors included; Z86.4 "personal history of psychoactive substance abuse" (recorded for 189 people), Z86.5 "personal history of other mental and behavioural disorders" (n ¼ 96), Z91.1 "personal history of noncompliance with medical treatment and regimen" (n ¼ 76), and Z91.5 "personal history of self-harm" (n ¼ 452).
Of particular interest in the context of this paper are those "factors influencing health status and contact with health services" which relate specifically to social determinants. In this database, a total of 43 references were made to social determinants of mental health problems (excluding substance use), for 39 individuals (0.2% of the whole dataset of 21,701 individuals, or 0.8% of the 4656 receiving a diagnosis).
To understand that in more detail, as listed in Tables 6-8, two people were identified as "unemployed" (Z56.0), and one was assigned the code Z56.6 for "other physical and mental strain related to work". One was identified as "homeless" (Z59.0) and three were classified as "living alone" (Z60.2). Two had an unspecified "problem Mental and behavioural disorders due to multiple drug use and use of other psychoactive substances: Psychotic disorder 9 0 Mixed and other personality disorders 2 0 F64.9 Gender identity disorder, unspecified 1 0 related to social environment" (Z60.9). Seven were coded as having "problems related to alleged sexual abuse of child by person within primary support group" (Z61.4) and two "problems related to alleged sexual abuse of child by person outside primary support group" (Z61.5). One person had "other negative life events in childhood" (Z61.8) and one unspecified "negative life events in childhood" (Z61.9). Six individuals had suffered the "disappearance and death of family member" (Z63.4), four "disruption of family by separation and divorce" (Z63.5), three "other stressful life events affecting family and household" (Z63.7) and one "other specified problems related to primary support group" Down's syndrome, unspecified 3 0  Problem related to social environment, unspecified 2 Z61.4 Problems related to alleged sexual abuse of child by person within primary support group 7 Z61.5 Problems related to alleged sexual abuse of child by person outside primary support group 2 Z61.8 Other negative life events in childhood 1 Z61.9 Negative life event in childhood, unspecified 1 Z63.4 Disappearance and death of family member 6 Z63.5 Disruption of family by separation and divorce 4 Z63.7 Other stressful life events affecting family and household 3 Z63.8 Other specified problems related to primary support group 1 Z65.3 Problems related to other legal circumstances 1 Z65.8 Other specified problems related to psychosocial circumstances 5 Z72.6 Gambling and betting 2 Z73.3 Stress, not elsewhere classified 1 Sub total 43 (Z63.8). One had reported "problems related to other legal circumstances" (Z65.3) and five "other specified problems related to psychosocial circumstances" (Z65.8). Two people were recorded as having problems with "gambling and betting" (Z72) one (noted earlier) was reported to be experiencing "stress, not elsewhere classified" (Z73.3).
In addition to Z-codes "factors influencing health status and contact with health services", the ICD system includes codes in chapter XX (20 in Roman numerals) related to "external causes of morbidity and mortality". These should be highly relevant to mental health, because they include sexual and other assaults, traumatic life events and such issues as war, terrorism and accidents. However, in this database, such available codes were again very rarely used. The "sequelae of motor-vehicle accident" (Y85.0) were recorded on 6 occasions, and "intentional self-harm by hanging, strangulation and suffocation" (X70.2) was recorded twice (with the slightly unexpected) "assault by pesticides" (X87.8) recorded once.

Social determinants for socially determined problems
In order to explore in more detail, the recognition given to social determinants, a breakdown by primary diagnosis (when given) was undertaken.
A primary diagnosis of "post-traumatic stress disorder" (F43.1) was given to 66 people. For those individuals, "other specified problems related to psychosocial circumstances" (Z65.8) was recorded on two occasions. No social Personal history of diseases of the nervous system and sense organs 12 Z86.7 Personal history of diseases of the circulatory system 62 Z87.0 Personal history of diseases of the respiratory system 2 Z87.1 Personal history of diseases of the digestive system 18 Z87.2 Personal history of diseases of the skin and subcutaneous tissue 1 Z87.3 Personal history of diseases of the musculoskeletal system and connective tissue 2 Z87.4 Personal history of diseases of the genitourinary system 9 Z87.5 Personal history of complications of pregnancy, childbirth and the puerperium 1 Z87. determinant was recorded for any of the other 64 people, although a "personal history of self-harm" (Z91.5) was recorded on five occasions, a "personal history of psychoactive substance abuse" (Z86.4) and "drug use" (Z72.2) were both recorded on one occasion, and a "personal history of noncompliance with medical treatment and regimen" (Z91.1) was recorded once (none of which are "social determinants" as defined by the World Health Organisation). This is a significant finding. The diagnosis of "post-traumatic stress disorder" quintessentially requires the identification of a traumatic event; the individual is distressed following, and consequent upon, a major life-event. And yet, here, the only codes that could have related to such events, were the two ambiguous references to "psychosocial circumstances". To be clear; no specific traumatic events were recorded for any of the persons diagnosed with PTSD. Similarly, 25 people had received a diagnosis of "adjustment disorder" (F43.2). Of these, seven people had received a code for a "personal history of self-harm" (Z91.5) and one for a "personal history of psychoactive substance abuse" (Z86.4); neither of which are social determinants, as defined by the World Health Organisation. No social determinant was therefore recorded; no traumatic events were listed. This is again significant, since "adjustment disorder" requires the diagnostic presence of a triggering event, which was not recorded in this database on any occasion.
Finally, 151 people had received a diagnosis of "emotionally unstable personality disorder" (F60.3). "Zcodes" were relatively common in this group, with 49 people (32.5%) receiving at least one code and 54 used in total. However, 42 of these codes referred to a "personal history of self-harm" (Z91.5), seven referred to a "personal history of psychoactive substance abuse" (Z86.4), four referred to "drug use" (Z72.2) and on only one occasion, to "problems related to alleged sexual abuse of child by person within primary support group" (Z61.4). This means that, excluding records of self-harm and substance use, in only one case out of 151 where a diagnosis of "emotionally unstable personality disorder" was used was a social determinantalleged sexual abuserecorded (0.7%).
The frequency of use of codes for these known social determinants was compared to estimates of population prevalence. Up to 39% of people receiving help for a mental health problem aged between 16 and 74 years are likely to be economically inactive (Singleton et al., 2003) and the rate of economic inactivity in the general public in the relevant city region in 2015 was 24.2% (Parkinson et al., 2016). In the current database, two individuals (0.009%) were identified as unemployed. Homelessness is, equally, a serious threat to mental health (Fazel et al., 2008), and in the catchment area for this study in 2015-2016 (the time-frame for this database) 312 decisions were made concerning the provision of accommodation following applications by homeless persons. One person in the current database was identified as homeless (0.005%). Three people (0.01%) were identified as living alone, and yet 11% of all mental health service-user in representative survey-study were living alone (Bijl & Ravelli, 2000). In the present database, adverse childhood experiences (Z61.4, n ¼ 7; Z61.5, n ¼ 2; Z61.8, n ¼ 1 and Z61.9, n ¼ 1) were recorded on 11 occasions (0.05%). This contrasts with 31% of participants reporting trauma events in childhood (Lewis et al., 2019) in epidemiological studies.

Discussion
Diagnostic information was present in only 4656 (21.5%) cases in this database of UK mental health care records from 2015. The remaining 17,045 (78.5%) cases included only data on the dates of episodes of care. Overall, codes for possible social determinants were used in only 39 cases (0.2% of the whole dataset of 21,701 individuals, or 0.8% of the 4656 receiving a primary diagnosis). Comparison with relevant baseline frequencies revealed a highly significant under-reporting of key, known, social determinants. Nineteen cases (0.1% of 21,701, or 0.4% of 4.656 cases with a mental health diagnosis) used codes for precise description of specific psychological phenomena. A close examination of both primary diagnosis and supplementary information also revealed a marked focus on patient's personal health behaviour, with substance use and self-harm frequently recorded, consistent with previous research which found that clinicians tend to record few relevant details of social determinants of health, commenting that the "social history" is "an underused tool" (Kenyon et al., 2007;p. 736).
The relatively low proportion of records including diagnostic information is unsurprising. Healthcare in the UK is provided universally, free at the point of use, and on the basis of the needs of the individual. Although there is an efficient record keeping, commissioning and procurement system, and employees are recompensed for their time; unlike in direct payment or insurance models of care, there is no need for clinicians to use diagnostic information in order, e.g. to bill for services. This finding immediately sheds doubt on a central set of claims for the necessity of medical diagnosis in mental healthcare; namely that it would be impossible to deliver services or clinical care in a nondiagnostic system (Frances, 2013). Notwithstanding the fact that many effective social services are non-medical and nondiagnostic, it is clear from the current study that diagnoses are not, in fact, central even to NHS mental health care.
Moreover, although the numbers were very small, the fact that the current database included cases for whom "primary diagnoses" were recorded using both codes for social determinants (e.g. Z73.3 "stress, not elsewhere classified") or phenomenological codes (R44 "auditory hallucinations") again, implies that alternatives to traditional medical diagnosis in mental healthcare are available and practitioners know about themthey are merely not frequently used.
Codes related to the social determinants of poor mental health were recorded for only 39 of the 21,701 recipients of care. This strikingly low use of available ICD codes is consistent with findings in the USA (Torres et al., 2017). Similarly, phenomenological codes were rarely recorded, in only 19 cases.
We do not know, from this dataset, what diagnostic information may have been individually held within psychiatric or nursing notes (textual information not necessarily equating to code entries, even if stored in electronic form), or in letters not included in the electronic database. It is likely that diagnostic information might have guided clinical decisions on at least some occasions, but was not recorded on the database. Nevertheless, the omission of records of the social determinants of mental health problems is important, because of the likely impact on our understanding of the problems, on care pathways, and on political agendas. Describing the circumstances that may have contributed to distress promotes understanding and thereby compassion (Jaspers, 1968). Research has demonstrated that the inclusion of information about social determinants reduces the likelihood that a pattern of behaviour is seen as pathological (Marsh et al., 2016) Omitting information about psychosocial circumstances means that a biomedical, pathologizing, explanation is more likely.
Comparison with known rates of incidence of social determinants of mental health problems (e.g. unemployment, homelessness, social isolation, or childhood trauma) suggests that clinicians are very significantly under-reporting these issues. As many professionals currently advocate a greater emphasis on prevention, including by politicians as well as clinicians (World Health Organization, 2013), omission of information about social determinants of mental ill health means that we are poorly armed with relevant data in the policy world. The information recorded in such registers or records shapes our understanding of mental health itself (Hess & Mendelsohn, 2010). Data on social and contextual factors are vital to public health approaches that see mental health as a community, not individualised issue (Clark & Gurewich, 2017;Whitehead & Dahlgren, 1991).