Bed occupancy 2: What factors affect it?

Introduction

In the first instalment of this series we looked at bed occupancy in general, and how it works in psychiatry. This blog will look in more detail at the factors which affect bed occupancy as they may be helpful in understanding the figures later on.

An overview of patient flow

Put simply, there are four key drivers that affect patient flow:

1. Demand;
2. Capacity;
3. Activity;
4. Queue.

These factors are commonly known as DCAQ and we will look at each in turn. Some of the definitions have been taken from the Scottish Government resources to healthcare providers.

1. Demand

Demand is the total sum of all requests into the service from all sources. In inpatient psychiatry this could include: Community Mental Health Teams (CMHTs); primary care (GPs); Crisis Teams; the Police; Crisis Hubs; and pretty much anywhere that might make a request for, or identify a need for, inpatient admission.

Gatekeeping and CRHTTs

It is not just those needing inpatient admission and it includes all those who may be perceived to need inpatient admission. This is important because much of the demand for inpatient care may not be the same population that inpatient psychiatrists think the service might be for. Most inpatient services do not have very explicit criteria for service provision, and there is fuzziness at the boundaries.

Let's assume that you have a limited number of inpatient beds and these are being 'gatekept' by a Crisis Response and Home Treatment Team (CRHTT). This is the model for CRHTTs that was developed many years ago (Department of Health, 2006). Since CRHTTs were funded from closing inpatient beds, they were considered to be essential for 'gatekeeping' to inpatient beds:

"It is important that mental health inpatient services and crisis services are joined up locally. It is necessary for a crisis team to act as gatekeeper for all people requiring access to inpatient services or other emergency care. Gatekeeping is an essential component of CRHT." (Department of Health, 2006)

Since they could offer home treatment as an alternative to admission, they could assess whether a patient was suitable for home treatment instead of an inpatient bed, thereby reducing demand for inpatient services.

The research on whether CRHTTs have achieved their core aim (reducing inpatient admission) is broadly supportive, but sometimes contradictory (for example: Barker, 2011; Jacobs, 2011; Stulz, 2020).

However, let's also assume that your CRHTT only has a locum consultant psychiatrist, many experienced staff have left, and it has seen what many CRHTTs in the UK have seen: their ability to provide home treatment has been compromised by the high demand for crisis assessment. This problem was recognised many years ago (Morgan, 2007) and the crisis response component of CRHTTs was recommended to be kept separate from the home treatment pathway, since the benefits of CRHTTs are only delivered when you can provide home treatment instead of hospital admission.

In these circumstances staff often find themselves simply trying to find a bed rather than delivering the higher-risk and more time-consuming process of home treatment. With regards to risk, it was recognised a long time ago that the suicide rate in CRHTTs was becoming higher than that in inpatient wards (Hunt, 2016). This was due to CRHTTs managing an increasingly risky population in line with their core purpose of being an alternative to hospital admission.

However, this increase in risk needs to be understood and managed by the service and without robust leadership and operational management, there is a likelihood that staff become wary of risk and will lower their threshold for admission. So, rather than preventing admissions to hospital it is possible for CRHTTs to increase admissions, since referrals that would not normally be considered for admission (and would receive home treatment) are now being admitted.

2. Capacity

Capacity is all the resources needed to do the work. In inpatient settings it consists of the buildings (i.e. beds and rooms) and also all the staff. If you close bedrooms (due to renovation, for example) or have staff off sick, then your capacity is reduced.

An inpatient bed is best understood as a combination of a physical bed and also the staff needed to provide care. Someone who needs a bed will also need nursing staff to provide care, and if their needs are more intense, they will need more staff. So, not all patients are equal but it is usually possible to average out admissions, beds, and staff to understand what your staffing requirements are.

Delayed discharges

Key things that can affect capacity include 'delayed discharges'. According to the Scottish Government (2016):

"A delayed discharge is a hospital inpatient who is clinically ready for discharge from inpatient hospital care and who continues to occupy a hospital bed beyond the ready for discharge date."

The higher the number of delayed discharges, the fewer beds will be available for new admissions. Reasons for a delayed discharge can include a lack of social care, or other factors needed that cannot be provided by the inpatient ward (such as renovation to someone's property, or white goods). In some cases, someone may be waiting for a nursing home or other specialist placement to be available and until it is, they cannot be discharged.

Delayed discharges in NHS Scotland are slowly increasing and are reported regularly by NHS Scotland. The most recent set of figures is shown below. The big drop in response to the COVID lockdown was created by large numbers of patients being discharged from hospital

Delayed discharges - NHS Scotland (April 2023)

3. Activity

Activity is the work done. In inpatient environments it consists of all the admissions, all the treatment, and the discharges (and everything in between). A number of things can affect overall activity. For example, a ward that has a higher turnover will have more activity: more people can be admitted and discharged.

Length of stay (LOS)

Often, higher turnover is due to a shorter length-of-stay (LOS). This means that more patients can be admitted and discharged in any given period of time. Whilst there is a concern that an insufficient LOS may increase the risk of readmission, in reality most wards will have a variation in LOS without seeing obvious differences in risk of readmission.

Local data would suggest that there is a range of LOS in which there will be no change in readmission, but as LOS gets very short the risk does appear to go up. Of course, LOS varies according to patient need but there are average values that can be used to understand how different wards are managing broadly similar patient groups.

Perverse incentives

It would be rare if such incentives did not exist within complex pathways and LOS is one example. It is easier to understand if we look at two wards, in the same hospital, and consider how individuals are incentivised to behave.

Ward A is led by a substantive consultant psychiatrist (Consultant A). The average LOS of Ward A is four days shorter than that of Ward B, but the risk of readmission after discharge is the same. The rate of incidents in Ward A is no higher than in Ward B so there is no indication that care is being compromised. Since the LOS is shorter, more patients are admitted and discharged.

Ward B is led by a locum consultant psychiatrist (Consultant B). The LOS is longer and fewer patients are admitted and discharged in any given time period. The Ward B consultant is more cautious about discharging patients because they are only a locum and they don't have as much experience.

Consultant A knows they are admitting and discharging more patients and has no hang-ups about not working as hard. They've seen the data and they know that their turnover indicates that they are working harder.

But they are also exposing themselves to more risk since it is very difficult (if not impossible) to predict which individual patients might have an adverse event (such as suicide) during their admission or shortly after discharge. If they admit and discharge more patients, their individual risk is greater as a result of seeing more patients. They also spend more time managing patients who are most unwell (since the total proportion of any admission when symptoms are greatest is higher with shorter admissions than with longer admissions).

The two wards are in the same hospital and managed by the same NHS Board. The notion of each ward admitting from a particular area (or group of CMHTs) has long gone and if a bed is available in either ward, the patient will be admitted there regardless of their GP, area, or CMHT.

Now let's imagine it's Thursday afternoon and both consultants are asked to discharge some patients to free up beds before the weekend. Consultant A knows that their LOS is already shorter than Consultant B and if they discharge patients from their own ward they will have the same number of new (and more unwell patients) tomorrow, and this will affect how much time they can spend with their existing patients. Meanwhile, Consultant B (by making no changes to their current practice) will have no new patients and will have a more predictable day.

Of course, it is the job of both consultants to admit and discharge patients and deliver optimum care in the safest way, but the incentives are not equal. Consultant A is incentivised to work less efficiently because they have already seen that this is a successful strategy for Consultant B.

Such perverse incentives will be familiar to most people in the NHS. If you work more efficiently than your colleagues, you usually get more work. Most high-performing doctors don't mind this most of the time because doing the work is rewarding and there were opportunities for teaching, research, and other meaningful activities. But the circumstances in which the NHS is currently working post-COVID (relentless demand, higher stress and burnout, real-world pay cuts over time, increases in moral injury, fewer opportunities for research) are different. Working in systems where unfairness is a daily lived experience will affect how people make choices about different priorities. If they didn't, we wouldn't be human.

4. Queue

Queue is also known as 'backlog' and is the activity that has not been dealt with. Typically, it will represent those patients who have been admitted elsewhere and need a bed locally, and also those patients who require a bed but have not had one available. They may have been becoming increasingly unwell in the community.

A queue is inevitable when demand exceeds capacity, and there will be work that has been displaced if inpatient services lack the capacity and activity to manage demand.

Queues can also form when activity is displaced. For example, if a CMHT has a high rate of staff absence or lacks consistent medical staff (For example: The Courier, 2023) then inpatient services may end up being the 'provider of last resort' and have to provide aspects of patient care that have not/ cannot be managed elsewhere in the system.

This brings us back to inpatient demand being a proxy for the functioning of all other parts of the mental health system.

References

Barker, V., Taylor, M., Kader, I., et al (2011) Impact of crisis resolution and home treatment services on user experience and admission to psychiatric hospital. The Psychiatrist, 35, 106-110. http://pb.rcpsych.org/cgi/content/abstract/35/3/106

Department of Health (2006) Guidance Statement on Fidelity and Best Practice for Crisis Services. London: HMSO. https://webarchive.nationalarchives.gov.uk/+/www.dh.gov.uk/assetRoot/04/14/16/73/04141673.pdf

Hunt, I. M., Appleby, L. & Kapur, N. (2016) Suicide under crisis resolution home treatment - a key setting for patient safety. The Psychiatrist, 40, 172-174. http://doi.org/10.1192/pb.bp.115.051227

Healey, D. (2023). EXCLUSIVE: Tayside psychiatry vacancies worst in Scotland as expert warns national services ‘unsafe’. The Courier, 24 January 2023. https://www.thecourier.co.uk/fp/politics/scottish-politics/4082341/tayside-psychiatry-scotland-unsafe/

Jacobs, R. & Barrenho, E. (2011) Impact of crisis resolution and home treatment teams on psychiatric admissions in England. British Journal of Psychiatry, 199, 71-76. http://dx.doi.org/10.1192/bjp.bp.110.079830

Morgan, S. (2007) Are Crisis Resolution & Home Treatment Services Seeing the Patients They Are Supposed To See? London: National Audit Office. http://www.nao.org.uk/idoc.ashx?docId=640def6f-837d-4f1e-a3e8-6a62194a30f1&version=-1

Stulz, N., Wyder, L., Maeck, L., et al (2020) Home treatment for acute mental healthcare: randomised controlled trial. British Journal of Psychiatry, 216, 323-330. https://doi.org/10.1192/bjp.2019.31

Bed occupancy 1: Why does it matter?

Introduction

This is part one of a series of blogs on bed occupancy in General Adult Psychiatry (GAP) in Scotland. In order to make sense of the findings of a Freedom of Information (Scotland) Act (FOISA) request to all NHS Boards, some background is probably necessary.

There are four parts:

1. An overview of bed occupancy in psychiatry, what we know about it, and why it's important.
2. The factors affecting bed occupancy.
3. Reporting of bed occupancy.
4. Bed occupancy in GAP in Scotland 2018-2023 - findings from a Freedom of Information request.

Why is bed occupancy an issue?

A colleague recently posted on twitter that there were no more psychiatric beds available in Scotland. For anyone who has been on-call for psychiatry in the last few years, this has seemed like an occasional occurrence. In the last year or two it does seem to be becoming more common although it is not clear if it is truly a lack of beds or whether the 'availability' of beds is dependent on who is asking.

For example, if you are an NHS Board that is usually running at over 100% occupancy, you are unlikely to be able to have beds in 24-48 hours, neighbouring NHS Boards will know this, and they may therefore assume that any patient you admit to them will not be transferred back quickly. All NHS Boards will usually keep the last 1-2 beds for their patients and they are reluctant to give up their last remaining beds when they may not get that bed back quickly.

The lack of beds has impacts for all parties involved.

For patients, it means that you might not get a local psychiatric bed and you may need to be transferred to another NHS Board. This is stressful, takes time, and you will receive care from a team and a service that you probably don't know. Your admission is likely to be longer because it is slower to deliver care when you don't have access to notes or those already involved with the patient.

For relatives, it means that your loved one may be admitted to another NHS Board area. There may not be public transport and if you have a car, you may need to spend time driving to visit your relative. This is an 'Out of Area Placement' (OAP) and the Nuffield Trust have recently published updated data on OAPs.

For staff, it means that you have to phone around neighbouring NHS Boards trying to find a bed. Often, they will tell you that they don't have a bed and you'll have to convince them that admission is the only option available to you. Usually, a consultant-to-consultant phone call is expected which means that  if the transfer if 'out-of-hours', two senior psychiatrists need to be woken up to have a discussion about a patient that neither has seen or probably will see. Meanwhile, the patient is sitting without a bed and waiting for transport.

Systems under stress

Although this series of blog posts relates primarily to bed availability, the number of available beds is a function of how well all the other parts of the system are working. For example, the Nuffield Trust have recognised that out-of-area placements arise due to a whole system under stress, saying:

"OAPs are an indicator of a whole mental health system under pressure, not simply the result of too few acute mental health beds. A lack of focus on prevention, high levels of delayed discharges, increasing pressures in community care, lack of crisis response, and a rise in Mental Health Act use can all increase the pressure on bed capacity, which can in turn lead to reliance on OAPs." (Nuffield Trust, 2022)

If your CMHTs are struggling, they are less able to take over the care of patients being discharged from hospital - this prolongs the length of stay. CMHTs may lack the staff to deliver more intensive care to patients who are becoming unwell, and the threshold for admission may reduce. Crisis Teams may be under-staffed and they may be unable to provide intensive home treatment to lots of patients. If you have high levels of locums (many of which may not be as experienced as a 'substantive' consultant psychiatrist) care may take longer to deliver.

Ultimately, all roads lead to the inpatient bed. It is the 'final common pathway' for a complex system of care that has been pushed to the limit and beyond it.

What is bed occupancy?

Put simply, bed occupancy is the proportion (expressed as a percentage) of your beds that are occupied at any given time (usually counted at midnight) by a patient. If you have twenty beds and you have admitted ten patients, you are 50% full. If you have admitted twenty patients, you are 100% full.

More on how bed occupancy is calculated and reported is in part 2 of this series.

Optimum bed occupancy

There is a general acceptance that the 'optimum' (or ideal) bed occupancy is 85%. The assumption is that this means that you are using your expensive inpatient beds most efficiently, with a low risk of overload/ failure, and that you also have some 'give' in the system for periods of higher demand.

This figure of 85% is widely used as a reasonable 'target' and has been for some considerable time. For example, the Royal College of Psychiatrists (RCPsych) stated in 1998 that:

"Bed occupancy should not exceed 85%, if a safe environment is to be provided. Higher rates of occupancy also lead to pressure for premature discharge, leading to disturbed behaviour in the community and early relapse." (RCPsych, 1998)

This College report is no longer available but in 2011 they continued to indicate that an occupancy figure of 85% or less was a marker of a 'good ward' (RCPsych, 2011). The RCPsych continues to argue that there should be a maximum bed occupancy of 85%.

NHS Providers also refers to an 85% target when reporting bed occupancy for the NHS in general. The British Medical Association (BMA) also use an 85% figure when reporting occupancy published by NHS England.

Problems with a 85% target

One problem with this 85% figure is that it doesn't automatically match the reality of health care and it doesn't guarantee that your load is being matched to need or that your care is safe and/ or effective. You can have unsafe care with low occupancy and safe care with high occupancy. It's arguably a proxy measure rather than a direct measure of good care and functioning systems.

It's not hard to recognise that you can struggle to provide good care with low occupancy (if you lack staff and/or if your patients are very unwell, for example) and in other cases you can cope with higher occupancy without being over-loaded (if most of your patients are simply waiting for discharge, for example).

A more detailed critique of this 85% figure and of the balance between optimised care and system load can be found on this Improvement Science Blog.

The risks of over-occupancy

Risks to patient care

There is extensive evidence that increased occupancy is associated with adverse outcomes. A study in 2011 reported that the loss of beds over time correlated with an increased risk of involuntary admissions (Keown, 2011). Further, there have also been numerous studies that have found consistent associations between over-occupancy and risk of violence and aggression in psychiatric wards (Palmstierna, 1991; El-Gilany, 2010; Grassi, 2006; Ng, 2001; Nijman, 1999; Palmstierna, 1995; Virtanen, 2011).

Risks to staff

It was always the case that there would be peaks of high occupancy but that after a week or two they would reduce and staff would have time to 'recover'. Over the last few years (and with the added impacts of the COVID-19 pandemic) it seems likely that constant high occupancy is taking its toll on staff. This occurs via the direct effects of high workload, but also due to the increased risks of adverse events and violence that occurs with over-occupancy.

When systems are under pressure it is always supervision, training, and recovery time (due to having to cover extra shifts) that are sacrificed first.

There is, therefore, compelling evidence that whatever the ideal occupancy level, if it's too high then there are negative effects on staff (stress and burnout) and patient outcomes (increased adverse events, typically aggression).

Managing bed occupancy (and pass beds)

Many people might think it's odd to have more patients than actual beds, but it is relatively common during periods of high demand. Whilst it's hard to admit more than one patient to a bed in other specialties, in psychiatry it is quite common.

Pass beds

Since it's routine for many patients to have passes as they approach discharge (particularly with longer admissions), there will be periods when their physical bed on the ward does not have someone in it. This is a 'pass bed' and it is frequently used to admit someone else in the short-term.

The hope/ plan is that you can then find another bed (or discharge the new patient) before the old one comes back. During this time, you will have a bed occupancy of over 100%. If you have twenty beds, all of which are full, but two patients out on pass, and two patients admitted to those pass beds, then your occupancy will be 110% (22/10 = 1.1 = 110%).

Problems arise if your old patient is not well enough to be discharged, or your new patient cannot be discharged quickly. You have essentially double-booked the bed. Although people will argue that they have an 'exit strategy' for double-booking the bed, often it doesn't work out that way and staff can spend hours juggling patients around.

Surge beds

At least one NHS Board has started using alternatives to actual beds in actual bedrooms. For example, NHS Tayside has been using 'surge' beds for some time. This means that there is a mattress in an emptied interview room that can be used to admit a patient, but that room does not have the same facilities as a standard bedroom. There is no en-suite bathroom, nowhere to store clothes or possessions, and it may not have the same safety features as a bona fide psychiatric bedroom.

It also means that if you are using your interview rooms as bedrooms, there are fewer rooms for staff to see patients. Routine activities take longer, and all aspects of patient care are affected.

The Mental Welfare Commission for Scotland (MWC) has acknowledged the use of these rooms and has commented:

"It offers very little space, comfort or privacy and no en-suite bathroom facilities. Apart from a single bed there is no other storage for a patient’s personal belongings. We raised our concerns about this policy and the compromises to patient care, treatment and dignity." (Mental Welfare Commission, 2022)

Why not just build more beds?

It may make sense to simply build more hospital beds. After all, this will give you the capacity you need won't it?

There are a number of issues, however:

1. Hospital beds are expensive. Most NHS Boards are struggling to fund existing services, let alone expand their current ones.
2. All beds have to be staffed. Even if you could build more beds you might not be able to find the staff to ensure that care is safe and effective. Nurses are leaving the profession (in all specialties) due to high stress and burnout. They are also realising that they can work with less stress and more money by being 'agency' nurses; often covering the gaps in the wards that they used to work in.

There is another good reason why building more beds doesn't always work: it's called Roemer's Law.

First proposed in 1959 (Shain, 1959) the principle states that: "in an insured population, a hospital bed built is a bed filled." Although there was some uncertainty about whether it only applied to specific situations (e.g. US-based healthcare), there is evidence that Roemer's Law appears robust (Delamater, 2013; Ginsburg, 1983; Phillip, 1984; Rohrer, 1990).

This means that you can't just build your way out of increased bed occupancy since your beds will almost certainly always be full however many you have. The second part of this series will look at the factors affecting bed occupancy and what options are available to address this issue.

References

Delamater, P. L., Messina, J. P., Grady, S. C., et al (2013) Do More Hospital Beds Lead to Higher Hospitalization Rates? A Spatial Examination of Roemer’s Law. PLoS One, 8, e54900. https://doi.org/10.1371/journal.pone.0054900

El-Gilany, A. H., El-Wehady, A. & Amr, M. (2010) Violence Against Primary Health Care Workers in Al-Hassa, Saudi Arabia. Journal of Interpersonal Violence, 25, 716-734. http://dx.doi.org/10.1177/0886260509334395

Ginsburg, P. B. & Koretz, D. M. (1983) Bed Availability and Hospital Utilization: Estimates of the "Roemer Effect". Health Care Financing Review, 5, 87-92. https://www.ncbi.nlm.nih.gov/pubmed/10310279

Grassi, L., Biancosino, B., Marmai, L., et al (2006) Violence in psychiatric units: a 7-year Italian study of persistently assaultive patients. Social Psychiatry and Psychiatric Epidemiology, 41, 698-703. http://dx.doi.org/10.1007/s00127-006-0088-5

Keown, P., Weich, S., Bhui, K. S., et al (2011) Association between provision of mental illness beds and rate of involuntary admissions in the NHS in England 1988-2008: ecological study. BMJ, 343, d3736. http://dx.doi.org/10.1136/bmj.d3736

Mental Welfare Commission for Scotland (2022) Report on announced visit to: Ward 2, Carseview Centre, 4 Tom Macdonald Avenue, Dundee DD2 1NH. Date of Visit: 29 November 2021. Edinburgh: Mental Welfare Commission for Scotland. https://www.mwcscot.org.uk/sites/default/files/2022-03/Carseview-Ward2-20211129-a.pdf

Ng, B., Kumar, S., Ranclaud, M., et al (2001) Ward Crowding and Incidents of Violence on an Acute Psychiatric Inpatient Unit. Psychiatric Services, 52, 521-525. http://ps.psychiatryonline.org/cgi/content/abstract/52/4/521 

Nijman, H. L. I. & Rector, G. (1999) Crowding and Aggression on Inpatient Psychiatric Wards. Psychiatric Services, 50, 830-831. http://ps.psychiatryonline.org/cgi/content/abstract/50/6/830 

Palmstierna, T. & Wistedt, B. (1995) Changes in the pattern of aggressive behaviour among inpatients with changed ward organization. Acta Psychiatrica Scandinavica, 91, 32-35. http://dx.doi.org/10.1111/j.1600-0447.1995.tb09738.x

Palmstierna, T., Huitfeldt, B. & Wistedt, B. (1991) The Relationship of Crowding and Aggressive Behavior on a Psychiatric Intensive Care Unit. Hospital & Community Psychiatry, 42, 1237-1240. http://dx.doi.org/10.1176/ps.42.12.1237

Phillip, P. J., Mullner, R. & Andes, S. (1984) Toward a better understanding of hospital occupancy rates. Health Care Financing Review, 5, 53-61. https://www.ncbi.nlm.nih.gov/pubmed/10310946

Rohrer, J. E. (1990) Supply-Induced Demand for Hospital Care. Health Services Management Research, 3, 41-48. https://doi.org/10.1177/095148489000300105

Royal College of Psychiatrists (1998) CR62. 'Not just Bricks and Mortar'. Report of the Working Group on the size, staffing, structure, siting and security of new acute adult psychiatric in-patient units. London: Royal College of Psychiatrists. http://www.rcpsych.ac.uk/publications/collegereports/cr/cr62.aspx [NO LONGER AVAILABLE]

Royal College of Psychiatrists (2011) OP79. Do the right thing: how to judge a good ward. Ten standards for adult in-patient mental healthcare. London: Royal College of Psychiatrists. http://www.rcpsych.ac.uk/usefulresources/publications/collegereports/op/op79.aspx [NO LONGER AVAILABLE]

Shain, M. & Roemer, M. I. (1959) Hospital costs relate to the supply of beds. Modern Hospital, 92, 71-73. https://www.ncbi.nlm.nih.gov/pubmed/13644010

Virtanen, M., Vahtera, J., Batty, G. D., et al (2011) Overcrowding in psychiatric wards and physical assaults on staff: data-linked longitudinal study. British Journal of Psychiatry, 198, 149-155. http://dx.doi.org/10.1192/bjp.bp.110.082388

Long-term outcomes in schizophrenia - can some people do without meds?

There's another study by Martin Harrow which has been published online by the Archives of General Psychiatry:

Harrow, M., Jobea, T. H. & Faulla, R. N. (2012) Do all schizophrenia patients need antipsychotic treatment continuously throughout their lifetime? A 20-year longitudinal study [In Press: doi:10.1017/S0033291712000220]. Psychological Medicine.


It will undoubtedly get a lot of attention because of some of the inferences that are being made. Essentially, it's a longer-term follow-up from a previous study:

Harrow, M. & Jobe, T. H. (2007) Factors involved in outcome and recovery in schizophrenia patients not on antipsychotic medications: a 15-year multifollow-up study. Journal of Nervous and Mental Disease, 195, 406-414.

It's being implied (here, for example) that the drugs might account for some of the differences in outcome between those who took antipsychotic drugs for many years and those who didn't. It's important to acknowledge that antipsychotic drugs may have adverse effects on outcome - what's important is that Martin Harrow's studies don't confirm this.

The 2012 study reports 20-year follow-ups of a cohort of patients that consisted of 139 who were diagnosed early in their illnesses. The diagnoses were: schizophrenia (N=61); schizoaffective disorder (N=9); and psychotic mood disorders (bipolar, N=38; unipolar, N=31). All patients were diagnosed according to DSM-III.

Patients have been followed-up at: 2, 4.5, 7.5, 10, 15 and 20 years after first hospitalisation. It is worth acknowledging that this kind of extensive follow-up is hard to do and rarely-done. In psychiatry, a one-year follow-up study is long-term. Harrow and Jobe (2012) have followed up 59 (84.3%) of the 70 patients with schizophrenia, and this group forms the basis of the most recent paper.

Some headline outcomes include:

• A greater proportion of patients who weren't on medication were in 'recovery' ("no positive or negative symptoms, and no rehospitalizations during the follow-up year") at each time point. See Figure 1, below.

• Those on antipsychotics had higher levels of anxiety at each time point. See Figure 2, below.

Of course, it's difficult to know whether there is a causal relationship. People on opioid painkillers are probably more likely to report pain than those who aren't, but we wouldn't necessarily suggest that the painkillers are the cause of the pain. We might reason that those with pain are more likely to need painkillers.

It is certainly possible that Martin Harrow is simply observing the same phenomenon - that those who have a better prognosis, and who have fewer symptoms over the years are least likely to be using antipsychotic medications. Indeed, in their 2007 paper they commented that: "The results suggest that the subgroup of schizophrenia patients not on medications was different in terms of being a self-selected group having better earlier prognostic and developmental potential." And in the 2012 paper they conclude that: "SZ patients not on antipsychotics for prolonged periods are a self-selected group with better internal resources associated with greater resiliency. They have better prognostic factors, better pre-morbid developmental achievements, less vulnerability to anxiety, better neurocognitive skills, less vulnerability to psychosis and experience more periods of recovery."

So, this is less about whether antipsychotics cause a worse outcome (since the study can't determine this), and more about the different paths that groups of patients take when they develop a schizophrenic illness.

I don't think that Martin Harrow is necessarily suggesting that the drugs influence outcome  per se, despite the fact that this is what is likely to be assumed by many readers.For example, the following are some comments from the Mad in America blog:

• "...those 'who were not on antipsychotic medications were significantly less psychotic than those on antipsychotics.'"

Again, we shouldn't be surprised. All this tells us that if you don't have psychotic symptoms, you probably don't need antipsychotic drugs. It doesn't mean that the drugs cause the symptoms. People who take inhalers for asthma are likely to have less good airways function, but it doesn't mean that inhalers reduce people's peak flow. The treatment with the drug is simply reflecting the need for ongoing treatment.

•  "This dramatic difference in anxiety symptoms remained throughout the 15 years, with more than half of those on antipsychotics still suffering from high anxiety at the end of 20 years."

Indeed, the differences in anxiety levels are notable (see Figure 2 above). However, it's hard to know what this is showing. Would anxiety levels (if due to drugs) take more than 2 years to develop? Possibly, but unlikely. Remember that differences between groups aren't that marked (in terms of recovery) at two years (see Figure 1 above), so all we can perhaps conclude is that the two different courses haven't separated out. When people start developing less favourable rates of recovery, their anxiety symptoms are higher than those patients who have greater recovery. It shouldn't be that surprising that for those with persistent symptoms, one of the symptoms is anxiety since this symptom is extremely common in most mental disorders.

• "At three of the six follow-ups, those off antipsychotics showed significantly better cognitive functioning, and in the other three follow-ups, there was a general trend favoring those off antipsychotics."

This isn't necessarily the case, and there are other ways of interpreting the data. The cognition data are as follows (Figure 3, below).

The data show that cognitive performance is relatively stable for those on meds, whilst it is highly variable for those not on meds. The differences between the groups aren't statistically significant between 2-10 years, and at 20-year follow-up. It's worth bearing in mind that one-in-six patients weren't assessed at 20-year follow-up, and it's possible that different patients are being tested at different time points; making reasoning about associations shaky.

What the Harrow long-term data do suggest is that some patients who have a diagnosis of psychotic illness don't necessarily need long-term antipsychotic drugs in order to have favourable outcomes, and many patients who don't take drugs do better than many patients who do. However, what the Harrow studies don't tell us is how to determine if a patient with newly-diagnosed schizophrenia will be in the good-outcome/ few drugs group. There are clues (discussed in the 2007 study cited above) regarding pre-morbid functioning and adjustment that are associated with a better prognosis but questions remain whether this can be translated into recommendations and confidence for specific patients.