A Short History of PROMs

This short history of patient-reported outcome measures is written to explain the origins of howRU, which has been particularly influenced by the work of Rachel Rosser.

The development of generic measures of health outcome has its origins in the application of operations research and systems thinking to healthcare evaluation and resource allocation during the late 1960s, in response to thinkers such as Donabedian, who popularised the distinction between the evaluation of Inputs, Process and Outcomes (Donabedian, 1966).

Input refers to the resources used or needed by a particular activity in terms of staff, facilities and consumables.  Input evaluation is typically carried out on the basis of comparative studies of indicators such as nurse/bed ratios where the nursing establishment and the number of beds are both input measures.

Process refers to the health care activity itself.  Process evaluation focuses on the quantity of activities performed and technical competence and thoroughness in carrying out those tasks.

Outcome refers to the consequences of health care activities in meeting objectives of treatment and patient progress.  Outcome evaluation is patient centred and addresses the impact of health service activities on patients’ quality of life, mortality and complication rates.

Health care outcome measures may be generic, meaning that they apply to all patients, or condition-specific, meaning that they are relevant for certain diagnostic groups.  Disadvantages of condition-specific measures are that they can not suitable for comparisons between disease categories, you need to know what is the matter with the patient before selecting the appropriate measure and they are ill-suited to patients with multiple problems. Here our attention is confined to generic measures of illness.

Typically, generic measures are made up of three separate components.  The first is a conceptual model relating to the underlying continuum of health, illness and death, the way morbidity and mortality may be combined in one measure, and the treatment of time.  Some generic measures seek to combine valuations of morbidity and mortality into a single measure, others treat these as separate types of outcome.  Our focus here is with measures that apply to living people.

The second is a classification of states of illness used to describe subjects (patients) such that each may be placed into one of a comprehensive and mutually exclusive set of categories.

The third component is a scale of valuations corresponding to the classification of states such that each category is attributed a valuation on a nominal, ordinal, interval or ratio scale (Stevens 1946).

A nominal scale is the weakest level of measurement and only distinguishes between mutually exclusive categories.  For example, the variables ‘sex’ (male and female) and ‘specialty (medicine, surgery etc.) are classified on nominal scales and no category has a greater value than any other. 

An ordinal scale, on the other hand, requires that classes can be ranked in order of value.  For example, a variable ‘pain/discomfort’ may have four classes (none, a little, a lot and extreme) and is measured at the ordinal level.  The classes can be ranked in the order given, but the difference between them is not known. 

In an interval scale this difference is quantified (so addition and subtraction are valid operations) whilst in a ratio scale there must be a true zero, and multiplication and division are permitted.  For example, patient age is measured on a ratio scale and it is valid to say that a patient aged 60 years is three times as old as one aged 20 (division) and that the difference in their ages is 40 years.  However, if age is grouped into broad age ranges (e.g. 20-29, 30-39, 40-49, 50-59 etc), the new variable ‘age group’ has ordinal properties. 

The Pioneers

During the late 1960s, several independent research groups in England, the United States and Canada began to undertake the development of generic outcome measures in which morbidity and mortality were combined into a single measure.

In the USA, Fanshel and Bush developed the concept of a health status index comprising an index of well-being and prognosis.  The output of a health programme is the change in the functional history of the target population resulting from the intervention of the health programme. The index of well-being represents the scaled valuations of a classification based on a conception of health in terms of function/dysfunction.  The state healthy (100% function) was given a value 1.0 and the state dead (absolute dysfunction) value 0.  All other states were assigned values between 0 and 1.0.  Methods used for scaling levels of well-being included category rating and magnitude estimation but these produced very different scales. Prognosis was defined as the transitional probability of change in functional state with time, but this was difficult to measure (Fanshel and Bush 1970). 

In Canada, Torrance developed a generalised cost-effectiveness model based, also using a function/dysfunction continuum from health (value 1.0) to dead (value 0).  Although the model is general the states of illness were drawn from particular diagnostic categories.  Methods used for scaling states of illness included the von Neumann-Morgenstern standard gamble and a time trade-off method, which has been widely adopted since (Torrance, Thomas and Sackett 1972)

At York, Williams and other health economists developed a conceptual framework of indicators of health status, need and effectiveness.  Some empirical work was done but they did not develop scales of valuation for health states. (Culyer, Lavers and Williams 1971)

In London, Rachel Rosser developed a scaled classification of states of illness with eight classes of disability and four classes of distress (Rosser and Watts 1972).

Disability relates to the extent that an individual is restricted in what he or she can or cannot do. The content and phrasing of Rosser’s disability classification has similarities with the Karnofsky (1948) scale, which is still used in cancer care.

Distress relates to the level of suffering in terms of pain, anxiety and depression that is experienced by each individual.  Distress must necessarily be inferred rather than observed and the classification of distress the assumption that unconscious patients are not in distress there are 29 categories of illness.

Rosser’s and Watts’ Classification of Disability and Distress

Disability

1.     No disability  

2.     Slight social disability

3.     Severe social disability and/or slight impairment of performance at work.  Able to do all housework except very heavy tasks.

4.     Choice of work or performance at work severely limited.  Housewives and old people able o do light housework only but able to go out shopping.

5.     Unable to undertake any paid employment.  Unable to continue any education.  Old people confined to home except for escorted outings and short walks and unable to do shopping.  Housewives only able to perform a few simple tasks.

6.     Confined to chair or to wheelchair or able to move around in the home only with support from an assistant.

7.     Confined to bed.

8.     Unconscious.

Distress

1.     No distress – no pain nor depression nor anxiety

2.     Mild - mild pin, e.g. slight toothache for which aspirin might be an effective treatment.  Understandable reactive states of mild depression.

3.     Moderate - moderate pain, e.g. severe toothache unrelieved by aspirin, severe migraine.  Moderate depression and anxiety for which medication would be prescribed.

4.     Severe  - severe pain, e.g. subarachnoid haemorrhage.  Suicidal   depression.  Overwhelming anxiety.

This classification was developed at Guy’s Hospital and was used in a study at St. Olave’s Hospital, which was part of the Guy’s group of hospitals.  This involved the classification of all patients on admission, on discharge, at the first outpatient appointment following discharge and on the first and last day of the one month study.  In all 48 doctors completed 2170 classifications.  The study demonstrated the feasibility of the method, its acceptability to medical staff and its sensitivity in demonstrating differences between units. 

Rosser’s and Watts’ classification of disability and distress offered a number of advantages over other patient classifications available during the 1970s and 80s:

  • Global: it has been applied successfully to patients to medical,  surgical and psychiatric wards from the age of 15 to 98.
  • Simple: it has only two dimensions, disability and distress; patients can easily be described in terms of code numbers (e.g. 7,2 confined to bed, mild distress).  Classifications having several dimensions cannot be used in this way.
  • Sensitive: there are 29 possible combinations of disability and distress (on the assumption that unconscious patients are not in distress), which discriminate between patients and show patient progress.
  • Reliable: has been demonstrated to be reliable where appropriate training has been given (Benson 1978).
  • Valid: the method had good clinical face validity.

The method was originally developed and used for measurement of the output of hospital units but it has been adapted to measure the population health status in terms of weighted life expectancy (QALYs) in a way that is similar to the proposals of Bush and Torrance.  It was never designed as a patient-reported measure.

The classification of disability and distress was scaled using the analysis of awards made by the courts to victims of accidents and using psychometric techniques based on extensive structured interview procedures with 70 subjects of different backgrounds (Rosser and Kind 1978).

Other groups developed general health profiles which introduced the patient as the rater, leading to the term patient-reported outcomes (PRO).  Early work on health profiles, such as the RAND Health Insurance Study (Ware 1976) and the Sickness Impact Profile (Bergner et  al 1976) used long questionnaires with over 100 items.

During the late 1980s, a second generation of data collection instruments emerged, which were designed to be completed by the patient for use in surveys and clinical research.  Many of these instruments are still in wide-spread use.

Rosser’s Index of Health-related Quality of Life

Later Rosser worked to develop a patient-reported measure known as the Index of Health-related Quality of Life (IHQL) This included a complex taxonomy with three top-level dimensions (Discomfort, Distress and Disability), seven attributes (Pain/discomfort, Symptoms, Dysphoria, Disharmony, Fulfilment, Dysfunction and Dependency), subdivided into 44 scales, with 107 descriptors and 225 descriptor levels.  The full structure is summarised below, to show the many different ways that illness can impact a person’s quality of life (Rosser et al 1992). 

Index of Health-related Quality of Life (IHQL)

  • Discomfort
    • Pain/discomfort: intensity, duration
    • Symptoms
      • breathlessness
      • disfigurement
      • fatigue
      • poor appetite
      • eating disturbance (overeating, bingeing, self-starving)
      • nausea
      • vomiting
      • sleep disturbance (too much, not rested, initial insomnia, interrupted, early waking)
  • Distress
    • Mood (dysphoria):
      • depressed
      • anxious
      • manic
      • distressing treatment
    • Disharmony:
      • social network (face to face, telephone, written)
      • communication (speech, writing, non-verbal)
      • intimacy (warm, confiding, listening, dependable, no close relationship)
      • sexual function (loss of interest, enjoyment, dysfunction)
    • Fulfillment:
      • feel blamed
      • stigmatized
      • socially unacceptable
      • loss of meaning
      • dissatisfied with life
      • bored
      • frustrated
      • a burden to others
      • loss of ambition
      • understand own state of health
      • informed
      • attitude to condition (fighting, resenting, accepting, with resignation, positively accepting, hopeless, denying, guilty, justly punished, disappointed),
      • perceived prognosis (cured, improved, the same, worse, fatal).
  • Disability
    • Dysfunction:
      • main occupation
      • other occupations
      • financial role
      • social role
      • family role
    • Dependence:
      • self-care (wash, dress feed)
      • mobility (walk, travel)
      • physical disability (paralysis, amputation, weakness, tremor, stiffness)
      • treatment away from home (general hospital, psychiatric unit, mental hospital, intensive care unit, hospice, hostel)
      • consciousness (unconscious, impaired consciousness)
      • cognitive function (memory loss, loss of concentration, confusion, disorientation in time, place or person)
      • body functions (incontinent bladder or bowel, dependent on machine inside or outside body, transplant dependent)
      • medicines/drugs/substances (prescribed, mind affecting, addictive)
      • sensory deficits (sight, hearing, taste, smell, touch).

It reminds me of the opening line of Tolstoy's Anna Karenina: “Happy families are all alike; every unhappy family is unhappy in its own way.” Illness affects people in many different ways.

The leading second generation instruments have been reviewed by McDowell (2006), and in a series of reports for the UK Department of Health for measuring the health of the general population (Sturgis et al 2001), long-term conditions (Fitzpatrick et al 2006) and elective surgery (Browne et al 2007).  The most widely used of these generic instruments are Short Form SF-36 (Ware and Sherbourne 1992) and its derivatives such as SF-12, the Euroqol EQ-5D (Brooks 1996) and the Health Utilities Index (HUI) (Horseman et al 2003).

References

Benson T. (1978) Classification of Disability and Distress by Ward Nurses: A Reliability Study. International Journal of Epidemiology, 7, 359-361.

Bergner M, Bobbitt RA, Kressel S, Pollard WE, Gilson BS and Morris JR. (1976) The sickness impact profile: conceptual formulation and methodology for the development of a health status measure. Int J Health Serv 6: 393-415.

Brooks RG. (1996) EuroQoL – the current state of play. Health Policy 37: 53-72.

Browne J et al. (2007) Patient reported outcome measures (PROMs) in elective surgery. Report for the Department of Health, London. (http://www.lshtm.ac.uk/hsru/research/PROMs-Report-12-Dec-07.pdf).

Culyer AJ, Lavers RJ and Williams A. (1971) Social Indicators: Health. Social Trends, 1971, 2: 31-42.

Donabedian A. (1966) Evaluating the Quality of Medical Care. The Millbank Memorial Fund Quarterly 44 (3): 166-203.

Fanshel S and Bush JW. (1970) A health status index and its application to health-services outcomes. Operations Research 18(6): 1021.

Fitzpatrick R et al. (2006) A structured review of patient-reported measures in relation to selected chronic conditions, perceptions of quality of care and carer impact. Patient-reported Health Instruments Group, Oxford.  Report to the Department of Health. (http://phi.uhce.ox.ac.uk/pdf/ChronicConditions/DH_REVIEWS_NOVEMBER_2007.pdf).

Horsman J, Furlong W, Feeny D and Torrance G. (2003) The Health Utility Index (HUI): concepts, measurement properties and applications.  Health and Quality of Life Outcomes 1:54. http://www.hqlo.com/content/1/1/54.

Karnofsky D, Abelmann W, Craver L, Burchenal J. The use of nitrogen mustards in the palliative treatment of carcinoma with particular reference to bronchogenic carcinoma. Cancer. 1948;1:634–56.

McDowell I. (2006). Measuring Health: A Guide to Rating Scales and Questionnaires. Third edition, Oxford University Press.

Rosser R and Kind P (1978).  A Scale of Valuation of States of Illness: Is there a Social Consensus?  International Journal of Epidemiology 1978, 7, 347-358.

Rosser R, Allison R, Butler C, Cottee M, Rabin R, Selai C. (1992) The index of health-related quality of life (IHQL): a new tool for audit and cost-per-QALY analysis. In: Walker SR, Rosser RM (editors). Quality of Life Assessment: Key Issues in the 1990s.  Kluwer Academic Publishers, pages 179-184 and 455-466.

Rosser RM and Watts VC. (1972) The measurement of hospital output. International Journal of Epidemiology, 1, 361-8.

Stevens, S.S (1946). On the Theory of Scales of Measurement. Science 103 (2684): 677–680.

Sturgis P, Thomas R, Purdon S, Bridgwood A and Dodd T. (2001) Comparative review and assessment of key health state measures of the general population. Department of Health, London.

Torrance GW, Thomas W and Sackett D. (1972) A Utility Maximisation Model for Evaluation of Health Care Programs. Health Serv Res. 7 (2) 118-133.

Ware JE Jr, Sherbourne CD. (1992) The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care 30:473-83.

Ware JE. (1976) Scales for Measuring General Health Perceptions," Health Services Research 11:396-415.