Aim
The aim of the economic evaluation was to compare the costs and benefits of a telephone advice and assessment service (referred to as ‘PhysioDirect’) with usual care in patients referred to the community physiotherapy service by primary care practitioners. Patients recruited to the study were randomly allocated to one of these two methods of service delivery. The primary analyses were from the perspectives of (1) the NHS and (2) patients and their families. Costs of lost productivity were also elicited and are presented separately.
The analysis was based on all costs incurred over the 6 months following randomisation to either PhysioDirect or usual care. Physical measures of resource use are presented separately from costs. The different viewpoints are separated throughout the analysis. For resource use, results are reported to two decimal places. For costs, results are reported to the nearest full penny. All costs are reported in 2009 prices.
The interventions
All patients were recruited following a GP referral to physiotherapy or patient self-referral.
Control Usual care involves an initial face-to-face assessment, which may be followed by one or more further appointments. Waiting times vary widely in different services and at different times but are typically between 4 and 8 weeks.
Intervention ‘PhysioDirect’ involves a telephone advice and computer-aided assessment, with written exercise and self-management information sent by post and face-to-face care if necessary.
The intervention and control arms of the trial are described in more detail in the previous chapter.
Form of analysis
Two forms of analysis were planned. The first used a cost consequences approach to compare the trial primary and secondary outcomes with cost from all three perspectives (NHS, patient and carer, and lost productivity). The second form of analysis planned was a cost-effectiveness (utility) analysis, with NHS costs being related to changes in QALYs. These were obtained using the EQ-5D measured at baseline, 6 weeks and 6 months, and valued using the tariff obtained from a UK general population survey to value the health states obtained using the EQ-5D measure.85 QALYs were estimated using the area-under-the-curve approach, adjusted for differences in baseline utility.86
Outcomes
The primary clinical outcome for the trial was the physical component summary (PCS) of the SF-36v2. Secondary outcomes included EQ-5D, SF-36v2 (MCS and individual scales), the MYMOP, personal perception of improvement, and patient satisfaction. Outcome data were collected by patient questionnaire at baseline, 6 weeks and 6 months.
Identification of relevant costs
The costs of the PhysioDirect service were compared with the costs of usual care from the point at which the patients were randomised to receive care in one of these two arms, until the 6-month follow-up. The analysis considered costs associated with the delivery of care from the perspective of the service provider, the patients and any carers, and the cost to society of lost production due to time off work by the patient and carers.
The scope of resource use identified as being relevant was identified in conjunction with physiotherapists and clinicians associated with the research and the provision of the two services. The analysis was confined to costs related to the reason for which the patient was referred to the physiotherapy service.
The costs identified as being of relevance in this analysis were as follows.
Direct costs to the health-care sector (NHS)
Cost of initial physiotherapy intervention, whether face-to-face or through PhysioDirect.
Cost of subsequent physiotherapy appointments, whether face-to-face or through PhysioDirect.
Primary and community care:
Face-to-face consultations GP, practice nurse, other health-care professionals.
Telephone consultations GP, practice nurse, other health-care professionals.
Out-of-hours contacts GP, practice nurse, other health-care professionals.
Home visits GP, community nursing teams.
Other primary care contacts NHS Direct, NHS walk-in centres.
Hospital care:
Patients and their carers
Travel costs associated with health-care visits including to the physiotherapy service.
Over-the-counter medication.
Expenditure on prescriptions.
Out-of-pocket expenditure associated with the purchase of private or alternative treatments, for example private physiotherapy, complementary and alternative therapies.
Cost of telephone calls relating to physiotherapy consultations.
Loss of earnings associated with lost employment for those unable to obtain sick pay (including the self-employed and those not entitled to sick pay).
Disability payments received. Disability payments are a cost from the perspective of the Personal Social Services (Government) but from the perspective of society are a transfer payment and are therefore ignored in the analysis conducted from the societal perspective.
Costs associated with lost production
Measurement of resource use
Much of the measurement of resource use came from computerised systems including the PhysioDirect specific system, the various community systems in operation across the study sites, and the various GP systems. Here data were collected on a per-patient basis using computerised data collection forms.
Further information about resource use, particularly in relation to patient resource use, was derived from the questionnaires that patients were asked to complete at the time of their initial referral to physiotherapy and at 6 weeks and 6 months following randomisation. Only costs associated with the condition for which the patient was receiving physiotherapy were included.
One other source of resource-use data was also used. Time and motion studies of the PhysioDirect service were used to obtain information about activities undertaken during patient non-contact time, a vital element in understanding the costs associated with PhysioDirect.
In all cases, the aim was to measure the resources used for the 6 months between randomisation and the final 6-month follow-up.
Direct costs to the health-care sector (NHS)
PhysioDirect consultations
Patient-level data were collected for each appointment with PhysioDirect. These included the length of the telephone call and the grade of the physiotherapist making that call. Data about the length of the telephone call were obtained by download from the PhysioDirect assessment software, which recorded which physiotherapist conducted each telephone call, and the duration of each call, based on when the physiotherapist logged in to and out of the patient's electronic record.
A small time and motion study comprising observation of sessions at each site was conducted to determine how physiotherapists occupied themselves when assigned to PhysioDirect but not on the telephone to patients. Observation of this non-contact time was required to estimate the opportunity cost of that time. Time was categorised as consultation time, administrative time related to PhysioDirect appointments, administrative time related to face-to-face appointments, general administrative activity (e.g. dealing with e-mails) and work breaks. The aim was to undertake observational time and motion studies at points in the study when sites were expected to be fully operational, across a mix of day, time of day and location. Observations were undertaken over a 4-month period (September to December 2009) of four PhysioDirect sessions in Bristol, three in Somerset and two in each of Stoke and Cheshire, with physiotherapists ranging from Agenda for Change Bands 6 to 8a (representing the full range of grades of staff providing PhysioDirect sessions).
During the time and motion study it was noted that physiotherapists had to continue some administrative activities following each telephone call, such as collating information to send to the patient by post. The time spent on these activities was not captured within the times logged in to the PhysioDirect software. However, in Bristol (but not in the other three physiotherapy sites), physiotherapists recorded details of telephone calls in the same routine records as they recorded face-to-face consultations. Comparison of matched data about individual calls demonstrated that the mean duration of calls recorded in the routine record was 5 minutes longer than that recorded in the PhysioDirect software. It was therefore assumed that this represented the extra administrative time spent after each call, and furthermore that this could be applied to calls in the other three sites. Consequently, 5 minutes was added to the duration of call time for each call. In a small number of instances we noted that physiotherapists had logged in to patients records for a very short time, sometimes several times in quick succession on the same day. We assumed this was due to problems telephoning a patient, or physiotherapists accessing the records for administrative purposes. These short ‘calls’ were included in the costing, but 5 minutes was only added once for any given patient on any one day, even if more than one call from the patient was recorded on the same day.
Information about the capital requirements for PhysioDirect was obtained through dialogue with the study sites and research teams and relevant costs were obtained through these routes. The capital requirements for each site varied but in general the teams needed far less space than for treating patients face to face. Based on the data collected, capital requirements were allocated at 50% of the capital requirements for face-to-face physiotherapy sessions utilised by Curtis.87
Face-to-face physiotherapy consultations
Data about face-to-face consultations were obtained from routine records maintained by physiotherapists at each site. These included the length of appointment, the grade of the physiotherapist seen and information about missed appointments. For most sites data were collected through the routine community databases. For Bristol and Somerset the ICS database was used; for Stoke the relevant database was Lorenzo. For Cheshire, where some sites did not use computerised systems, data were extracted from the patients' written notes.
Primary care consultations
Information was obtained from general practice records by researchers using a standard pro forma. Relevant consultations were identified as follows:
If the MSK condition for which the patient was referred was mentioned, the consultation was recorded, even if this was not the primary reason for the consultation.
If a patient had a MSK condition, but had a consultation about other problems and the MSK condition was not mentioned, these consultations were not recorded.
If a repeat prescription was issued on the same date as a consultation (i.e. was probably issued at the consultation) but there was no mention of the MSK condition at the consultation then the consultation was not recorded (but the prescription was recorded along with other repeat prescriptions for MSK problems).
Primary care consultations were recorded by type (e.g. face to face, telephone, out of hours, home) and by type of professional seen (e.g. GP, nurse).
Prescribed medication
Information was obtained from practice records by researchers using a standard pro forma. Relevant medication was defined as that prescribed for MSK conditions. It was not feasible to distinguish between medication prescribed for the condition for which the patient was referred for physiotherapy and any other MSK problem and so all MSK medication prescribed was included.
Any prescribed item that appeared in the chapters of the British National Formulary (BNF) detailed below was assumed to be a relevant prescription and details were recorded:
4.7.1–4.7.2 Analgesics
10.1.1 Non-steroidal anti-inflammatory drugs
10.1.2.2 Local corticosteroid injections
10.3 Drugs for the relief of soft-tissue inflammation.
Medication was recorded by name, route of administration, quantity and strength.
NHS Direct/walk-in centre consultations
Information about visits to NHS Direct or to walk-in centres related to the condition for which the patient was referred for physiotherapy was collected through the 6-week and 6-month questionnaires.
Visits to accident and emergency
Information about visits to A&E services was obtained from the questionnaires administered at 6 weeks and 6 months.
Secondary care outpatient consultations
Information about secondary care outpatient consultations related to the condition for which the patient was referred to the physiotherapy service was obtained from the questionnaires administered at 6 weeks and 6 months. This included information about the clinic name or department visited and the number of appointments. Where questionnaire data were ambiguous, for example with regard to the specialty, information about these consultations was obtained from GP records.
Secondary care inpatient stays
Information about secondary care inpatient stays related to the condition for which the patient was referred to the physiotherapy service was obtained from the questionnaires administered at 6 weeks and 6 months. Information obtained directly from this questionnaire was just whether or not the patient had received an inpatient stay. Any patient who answered ‘yes’ to this question was then telephoned to obtain further details about the location and length of stay.
Patients and their carers
Travel costs
Information about travel costs associated with health-care consultations for the condition being treated by physiotherapy was obtained from the patient questionnaires. Costs associated with travel to the GP and to the physiotherapy services were collected separately and then applied to all appropriate consultations.
Over-the-counter medication
Information about over-the-counter medication associated with the condition being treated by physiotherapy was obtained from the patient questionnaires. Individuals were asked to specify whether or not they had purchased over-the-counter medication and the total costs associated with any such purchases. They were not asked for details about the type of medication taken or the dosage.
Expenditure on prescriptions
Information about whether or not individuals paid for prescriptions was collected through the questionnaire. Individuals were asked to specify whether or not they paid for prescriptions, and if so, whether they paid on a per-item basis, or by purchasing a 3- or 12-month prepayment certificate.
Out-of-pocket expenditure
Information about out-of-pocket expenditure associated with the condition being treated by physiotherapy was obtained from the patient questionnaires. Individuals were asked to specify whether or not they had purchased any equipment or devices for the condition, or had had to pay for any extra help in the home (e.g. cleaning, gardening, ironing) or had paid for any form of private care (e.g. private physiotherapy, chiropractor, osteopath, other complementary and alternative therapies).
Loss of earnings
Information about whether or not the individual had experienced any loss of earnings associated with their condition was obtained directly from the individual through the patient questionnaires.
Disability payments
Information about disability payments received as a result of the condition for which the physiotherapy was provided was obtained from the questionnaires administered at 6 weeks and 6 months. This included information about the type(s) of benefit(s) received and the total amount paid over the previous 6 weeks (for the 6-week questionnaire) and since the previous questionnaire (for the 6-month questionnaire).
Costs associated with lost production
Information about whether the individual had experienced loss of productivity associated with their condition was obtained through the patient questionnaires. A number of questions were asked of each individual to facilitate the most accurate costing possible without asking questions that were excessively burdensome to respondents. Individuals were first asked whether or not they were in paid work. Individuals responding positively to this question were then asked whether or not their employment had been affected by their condition, specifically in terms of whether they had had to take sick leave, to work reduced hours or to take on restricted or altered activities. The total amount of time lost from work was asked. Similar information was also requested for time spent attending physiotherapy and general practice appointments, including telephone consultations with PhysioDirect.
Dealing with missing resource-use data
The extraction of computerised data from the physiotherapy records resulted in no missing data. Inevitably, however, some questionnaire and GP record data were missing. These missing data potentially affect all analyses, as elements of analysis from each perspective drew on questionnaire data for at least some aspects of resource use and on the questionnaire for analysis of outcomes in terms of QALYs gained.
Questionnaire data were missing for three reasons:
Patients withdrew from the study thus failing to complete questionnaires subsequent to their withdrawal.
Patients failed to complete entire questionnaires.
Patients failed to complete particular items within questionnaires.
General practitioner record data were missing for two reasons:
Patients moved practice during the trial period.
Patients records could not be identified at general practices.
For the baseline analysis, no attempt has been made to impute missing data arising from any of these causes. Within the sensitivity analyses, however (see below), missing data are imputed.
Valuation of resource use
Direct costs to the health-care sector (NHS)
PhysioDirect consultations
Methods used by Curtis87 in her estimation of a face-to-face physiotherapy appointment for a Band 5 member of staff were used in developing detailed costs for PhysioDirect appointments for each band of staff and in each location, based on median pay rates for each band from 1 April 2009 using the Agenda for Change.88 Data on national insurance payments, overheads and capital overheads were taken from Curtis.87 Capital overheads for telephone consultations were adjusted to 50% of the capital overheads for a face-to-face appointment to reflect the smaller amount of space required. Assumptions about the length of the working week and the number of weeks worked per year were also taken from Curtis.87 More detailed information about salary and the associated superannuation costs was then utilised to obtain a cost for each band of staff from 3 to 8a.
In order to obtain a cost per hour of telephone contact, we combined the Curtis87 estimates of cost per hour with information from the computerised records of the PhysioDirect service once it had reached a steady state of operation. These records identified the proportion of time spent by physiotherapists actually dealing with PhysioDirect patients. To this was added data from the time and motion study, which identified activities undertaken during non-contact time, for example administration for face-to-face patients (rather than PhysioDirect patients) or general administration. All estimates were carried out on a site-specific basis.
Face-to-face physiotherapy consultations
Methods used by Curtis87 in her estimation of a face-to-face physiotherapy appointment for a Band 5 member of staff were used to develop more detailed costs for physiotherapy appointments for each band of staff and in each location, based on median pay rates for each band from 1 April 2009 using the Agenda for Change.88 Data on national insurance payments, overheads and capital overheads were taken from Curtis. Assumptions about the length of the working week and the number of weeks worked per year were also taken from Curtis.87 More detailed information about salary and the associated superannuation costs was then utilised to obtain a cost for each band of staff from 3 to 8a. Finally, information obtained from physiotherapy service managers about the level of non-contact time in each study site for each band of staff was used to obtain a site-specific cost for each band of staff.
Primary care consultations
Information about the costs associated with primary care consultations was primarily obtained from Unit Costs of Health and Social Care (Curtis87) on a per-surgery consultation or home visit basis as appropriate. Costs used are detailed in . Costs excluding qualification are used for all visits, as this service is not likely to have implications for the numbers or training of primary care staff. Costs for out-of-hours appointments with a GP were obtained using recent publications.89
Values used for primary care consultations
Prescribed medication
The value of prescriptions issued in general practice (both when prescribed during a consultation or as a repeat prescription) was based on an estimate of the full cost to the NHS and was made up of four elements:
The basic price of the drug – from BNF
90 – to cost the prescribed medication.
An average deduction for discount. This depends on the total value of prescriptions dispensed from the pharmacy in a month and varies from a 5.63% deduction to a 11.5% deduction (May 2010). A mid-point deduction of 8.56% was applied to all basic drug prices.
The professional fee: 90p per prescription (obtained from the Drug Tariff England and Wales: www.nhsbsa.nhs.uk/924.aspx).
The container allowance: 3.24p per prescription (obtained from the Drug Tariff England and Wales).
NHS walk-in-centre consultations
The national evaluation was used to value a consultation at a walk-in centre,91 with values adjusted to 2009 values using the pay and prices index.87 The consultation value used was £31.99.
Visits to accident and emergency
Information about the costs of visits to A&E services was obtained from the National Schedule of Reference Costs for 2009–2010, using the value for A&E attendances that do not lead to subsequent admission.92 The value used for an A&E consultation was £103.
Secondary care outpatient consultations and procedures
Information about the costs of visits to outpatient consultations in secondary care was largely obtained from the National Schedule of Reference Costs for 2009–2010,92 supplemented by direct information from the Bristol Homeopathic Hospital. Values used are shown in for those specialties that were utilised by patients included in the trial.
Values used for outpatient consultations
Information about the costs of procedures conducted in outpatient departments in secondary care was primarily obtained from the same source.92 Values used are shown in for those procedures that were utilised by patients included in the trial.
Values used for procedures undertaken in outpatients
Secondary care inpatient stays
Information about the costs of inpatient hospital stays was based upon the healthcare resource group (HRG) for the particular admission. HRGs were coded using an earlier version of HRGs than the current version, because of the fuller information available, and so cost data were used from 2007 to 2008 and inflated to 2009 prices. Costs were obtained from the National Schedule of Reference Costs 2007–2008 (www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/DH_098945). Values used are shown in .
Values used for inpatient hospital stay
Patients and their carers
Cost of telephone calls to the physiotherapy service
The cost of telephone calls to the physiotherapy service was borne by patients. These were costed at the standard British Telecom rate prevailing at the time, i.e. 5.25p per minute plus 9p connection charge. Patients using the face-to-face service are required to call to make an appointment; in these cases a nominal charge relating to a 3-minute call was included.
It is recognised that patients may have used mobile phones on a ‘pay as you go’ basis, making calls more expensive, or may have had contract mobile phones or belonged to landline telephone contracts with included ‘free’ minutes. However, no information was available about these factors and so the standard British Telecom call rate was used.
Travel costs
Information about travel costs was obtained directly from the patient questionnaires except for the case of car mileage. Information about mileage costs was obtained from the Automobile Association schedule and a value of 46p per mile was used.
Over-the-counter medication
The total cost associated with over-the-counter medication was recorded by the individual in the questionnaire.
Expenditure on prescriptions
Information about the cost of prescription charges was obtained from the Department of Health.
Out-of-pocket expenditure
Information about the costs of out-of-pocket expenditure was obtained directly from the patient through either the initial questionnaire or the supplemental telephone interview.
Loss of earnings
Information about direct loss of earnings experienced by patients because of their condition was obtained through the questionnaire.
Disability payments
Costs associated with disability payments were obtained directly from the patient questionnaire.
Costs associated with lost production
It was assumed that all absence from work represented a loss in productivity and the human capital approach was therefore used to assign costs. Median hourly earnings (excluding overtime) by age and sex were obtained from the Office of National Statistics 2009 annual survey of hours and earnings,94 and applied to all estimates of time lost from work ().
Weekly salary costs (£) used for lost productivity by age and sex
Discounting
Costs and outcomes were not discounted, as the study was limited to a period of 6 months.
Uncertainty
There were three elements involved in addressing uncertainty within the analysis.
Uncertainty resulting from patient variation
Uncertainty in the cost-effectiveness–utility ratios resulting from patient variation in resource use and effectiveness was captured by estimating CIs around the net benefit statistic and estimating cost-effectiveness acceptability curves (CEACs).
Uncertainty in specific estimates
In the trial, the same level of physiotherapist staffing was available to patients in the PhysioDirect and usual-care arms. The amount of staff time devoted to PhysioDirect telephone sessions was based on estimates of the likely proportions of patients being managed by telephone or face-to-face consultations. There is good reason to believe that because of the constraints of the randomised trial design, the PhysioDirect service did not operate at full capacity. Sensitivity analysis was therefore used to estimate relative cost-effectiveness assuming a more efficient PhysioDirect service. In two of the physiotherapy services participating in the trial the PhysioDirect service continued after the trial had ended and the physiotherapists' time was used more productively for several reasons. First, in the main trial only a minority of patients received PhysioDirect because of the combined effect of exclusions, non-participation in the trial or being randomised to usual care. Outside the trial the patient throughput to PhysioDirect was much higher and so the physiotherapists were busier. Second, staffing levels were adjusted in the light of experience to try to ensure that there were enough staff to operate the service but be working most of the time. Third, instead of trying to staff the service so that patients could be connected to a physiotherapist as soon as they phoned, after the trial the service mainly operated a call-back system in which patients telephoned and left their contact details and then the physiotherapists called them back at an agreed time. Data from one service (Bristol) on staffing levels and all call times from January to March 2011 were used as a basis for estimating the cost of running the service as if these conditions had prevailed during the trial. The data about staffing levels, number of calls and call durations were obtained in the same way as in the trial.
Sensitivity analysis was also carried out to test the effect of including or excluding hospital costs. Typically, in a primary care trial, few patients use secondary care but the cost, for those who do, is very high, and so small differences between trial arms in the number of people using secondary care may have a disproportionate effect.
Uncertainty resulting from missing data
The entire analysis was rerun following multiple imputation of missing data. The multiple imputation by chained equation procedure95 was used. This method uses regression techniques based on the values of available data to estimate missing values. Variables included in the regression model are those likely to predict most closely the missing values. The regression model used to impute missing cost data included age, sex, randomisation group, GP costs, practice nurse costs, cost of other primary care professionals, cost of prescribed medication and hospital costs. Missing EQ-5D scores were imputed using data on age, sex, randomisation group, PCT, referral problem, and SF-36v2 PCS and EQ-5D at each time point.
