We have a new website go to gov.scot

The Scottish Executive Central Heating Programme: Assessing Impacts on Health

Listen

APPENDIX B ANALYTICAL METHODS

B.1 Introduction

As outlined in the Summary of Findings ( Section 1), a total of 67 individual outcome measures was investigated as part of the evaluation. Analysis of these was conducted either by analysis of covariance (for continuous outcomes measures) or by logistic regression (in the case of binary outcomes). For each measure, a statistical model was constructed in which the outcome at the endpoint of the evaluation was predicted by:-

a) The value of the outcome at the initial point ( i.e. the first interview with the respondent)

b) A binary indicator representing whether the respondent was a member of the comparison group, or was a recipient of central heating under the Programme

c) A set of additional covariates representing factors (such as the age and gender of the respondent) which could potentially influence any relationship between the outcome and the receipt of central heating

Of the above, item (a) was included to guard against the results being influenced by the statistical phenomenon known as 'regression to the mean' (simply stated, the tendency of 'extreme' values to become less extreme over time irrespective of specific actions or influences which may be in operation e.g. the provision of new heating systems). Item (b) is the main result of interest, for it is the values of this - obtained via the statistical models - which permit a determination of whether households which received heating under the Programme report different values of the outcomes from those of the comparison group. The additional factors (Item [c]) were included to ensure that, so far as is practically feasible, the influence of (for example) the age and gender of the respondent were adjusted for when estimating the effect of the Programme on a specific outcome measure. The precise set of such factors or predictors included in the statistical modelling process varied according to the outcome of interest. For example, an outcome representing some measure of health required adjustment for the possible influence of tobacco smoke exposure, while a measure representing heating usage did not.

All analyses were weighted - via the method of inverse propensity scores [ref. 6] - to adjust for attrition of respondents between the initial and final interviews.

The following material provides, for each outcome measure:-

a) A brief account of how the measure was defined

b) The type of model used (either analysis of covariance or logistic regression), and an indication of which additional predictors were featured in the model. Three standard sets of predictors were used (referred to respectively as Predictor set A, B and C). A full listing of the elements included in each set is given later in this Appendix (Section B.13). For a small number of outcomes, technical considerations meant that none of the standard predictor sets could be used satisfactorily. In such cases, the actual predictors used are explicitly specified.

Sections B.2 to B.12 give the information outlined above ( i.e. items [a] to [c]) for the individual conceptual areas featured in the analysis, and thus correspond to Sections 4.2 to 4.12 in the body of the Report.

B.2 Perceptions of warmth in the home

B.2.1 Whether respondent is kept sufficiently warm by heating during the period October - March

Definition

This measure is derived from responses to the question "During the colder months (October - March), do you generally find that your heating keeps you warm enough at home or not?". The available responses to the question are: "Yes, always", "Only some of the time" and "No, never". For the analysis responses were dichotomised, contrasting "Yes, always" against the two remaining (less favourable) responses.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.2.2 Whether inadequate heating presents a serious problem

Definition

This measure is derived from a supplementary question to that cited above, the additional prompt to respondents taking the form "How much of a problem is this, if at all, to you?". The supplementary item is presented only to those who give a reply of other than "Yes, always" to the preceding question, and offers the following response options: "A serious problem", "A bit of a problem", "Not very much of a problem" and "Not a problem". For analysis purposes, responses were dichotomised such as to contrast "A serious problem" against the three remaining (less severe) responses.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.2.3 Whether respondent is satisfied with heating

Definition

The respondent's level of satisfaction with her / his heating is elicited via the question "Overall, how satisfied or dissatisfied are you with your heating?". The response options provided are: "Very satisfied", "Fairly satisfied", "Neither satisfied nor dissatisfied", "Fairly dissatisfied" and "Very dissatisfied". To perform the analysis, responses were dichotomised such as to contrast "Very satisfied" OR "Fairly satisfied" against the remaining responses (which indicate lower levels of satisfaction).

Type of model / Additional predictors

Logistic regression / Predictor set C

B.3 Patterns of heating in the home

B.3.1 Whether more than half of the rooms in the home are permanently unheated in cold weather

Definition

This measure is derived from responses to a linked group of questions asking for how long each individual room in the home ( e.g. kitchen, bathroom) is heated in winter. Responses are collected via a system of duration bands, the available options being "0 (no hours)", "1-3 hours", "4-6 hours", "7-9 hours", "10-12 hours", "13-16 hours", "17-19 hours", "20-23 hours" and "24 hours". To derive the measure, the number of rooms identified as being heated for zero hours is divided by the total number of rooms in the dwelling. The result of this calculation is then reduced to a binary quantity, contrasting 0.5 (half) of the total number of rooms or less against more than half of the total number of rooms. The final value thus represents whether more than half of the rooms in the home are unheated in winter.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.3.2 Whether more than half of the rooms in the home are permanently heated (24 hours per day) in cold weather

Definition

This measure is derived as described in B.3.1 above, except that the numerator in the calculation is the number of rooms reported as being heated for 24 hours per day. The final value of this measure represents whether more than half of the rooms in the home are permanently heated in winter.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.3.3 Whether more than half of the rooms in the home are heated for nine hours per day or less in cold weather

Definition

This measure is derived as described in B.3.1 above, except that the numerator in the calculation is the number of rooms reported as being heated for any of "0 (no hours)", "1-3 hours", "4-6 hours" or "7-9 hours". The final value of this measure represents whether more than half of the rooms in the home are heated for nine hours per day or less in winter.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.3.4 Duration (hours per day) for which the dwelling's kitchen is heated in cold weather

Definition

This measure is derived from responses to the appropriate item in the set of questions used in B.3.1 - B.3.3 above. To arrive at a numeric value appropriate for use in analysis, the banded representation of heating duration (see B.3.1) is replaced by the mid-point of each band. Thus, an original response of "4-6 hours" is converted to a notional value of 5 hours; "13-16 hours" becomes 14.5 hours; and so on.

Type of model / Additional predictors

Analysis of covariance / Predictor set C

B.3.5 Duration (hours per day) for which the dwelling's bathroom is heated in cold weather

Definition, type of model and additional predictors as B.3.4.

B.3.6 Duration (hours per day) for which the dwelling's main living room is heated in cold weather

As B.3.4.

B.3.7 Duration (hours per day) for which the dwelling's hall is heated in cold weather

As B.3.4.

B.3.8 Duration (hours per day) for which the dwelling's main bedroom is heated in cold weather

As B.3.4.

B.3.9 Duration (hours per day) for which the dwelling's second bedroom is heated in cold weather

As B.3.4.

B.3.10 Average duration (hours per day) for which rooms in the dwelling are heated

Definition

This measure is derived by summing the heating duration values reported for individual rooms ( i.e. the values derived at B.3.4 to B.3.9 above), then dividing this sum by the number of rooms in the dwelling.

Type of model / Additional predictors

Analysis of covariance / Predictor set C

B.4 Condensation, dampness and mould

B.4.1 Whether use of any rooms in the home is avoided due to difficulty in heating

Definition

This measure is derived from responses to the question "Are there any rooms in your home that you avoid using because you have difficulty heating them?". The available response options offer a simple binary choice ("Yes" and "No, none"). The analysis models the probability of a "Yes" response.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.4.2 Whether any environmental problem (condensation / damp /mould) is present in the kitchen

Definition

This measure is derived from responses to a series of questions asking which of a range of environmental problems are experienced in individual rooms within the dwelling. The problems specified are "Condensation on windows or walls", "Damp smell", "Mould growth on carpets / curtains / furniture", "Mould growth on walls, ceilings or floors", "Mould or rot in window frames" and "Other problems". The information is elicited separately for each room in the home (kitchen, bathroom etc.), and respondents are permitted to report multiple problems within a room where appropriate. The measure is derived by determining whether one or more of the specific problem types ( i.e. excluding "Other problems") is identified in the kitchen.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.4.3 Whether any environmental problem (condensation / damp /mould) is present in the bathroom

Definition, type of model and additional predictors as B.4.2

B.4.4 Whether any environmental problem (condensation / damp /mould) is present in the main living room

As B.4.2.

B.4.5 Whether any environmental problem (condensation / damp /mould) is present in the hall

As B.4.2.

B.4.6 Whether any environmental problem (condensation / damp /mould) is present in the main bedroom

As B.4.2.

B.4.7 Whether any environmental problem (condensation / damp /mould) is present in the second bedroom

As B.4.2.

B.4.8 Whether any environmental problems (condensation / damp /mould) cause serious difficulty

Definition

This measure is derived only for those respondents who report the presence of any environmental problems in the home. Such respondents are presented with the question "Overall, how much do these problems cause you difficulty in daily life?". The response options offered are "A lot of difficulty", "A little difficulty" and "No difficulty at all". For the analysis responses were reduced to a binary quantity, contrasting "A lot of difficulty" with the two remaining options. The resultant measure represents whether environmental problems in the home are perceived as causing serious difficulty.

Type of model

Logistic regression

Additional predictors

Technical considerations (specifically, sparseness in the data) meant that the standard predictor set C could not be used for this analysis, predictor set B being used instead.

B.4.9 Whether any rooms in the home cannot be used due to problems of damp or condensation

Definition

As with B.4.8, this measure is derived only for those respondents who report the presence of any environmental problems in the home. Such respondents are presented with the question "Are there any rooms in your home that you are unable to use because of problems of damp or condensation?". The response options provide a simple binary choice between "Yes" and "No, none". The analysis models the probability of a "Yes" response.

Type of model

Logistic regression

Additional predictors

Technical considerations (specifically, sparseness in the data) meant that the standard predictor set C could not be used for this analysis, additional predictors being restricted to age, gender and housing tenure.

B.5 Overall satisfaction with the home

B.5.1 Respondent's overall satisfaction with her / his home

Definition

This measure is based on responses to a questionnaire item which asks "On the whole, how satisfied or dissatisfied are you with this house or flat?". The response options offered are "Very satisfied", "Fairly satisfied", "No opinion", "Fairly dissatisfied" and "Very dissatisfied". The original responses were reduced to a binary quantity by contrasting "Very satisfied" against the remaining responses.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.5.2 Respondent's perception of home as "A place I want to get away from"

Definition

This measure is derived from responses to a questionnaire item which elicits the respondent's level of agreement with the statement "My house / flat is a place I want to get away from". The response options offered are "Strongly agree", "Agree", "Neither agree nor disagree", "Disagree" and "Strongly disagree". For the analysis, responses were dichotomised such as to contrast "Strongly disagree" (the most favourable response i.e. that indicating the greatest degree of attachment to the home) against all other responses. This approach creates a measure which represents whether the respondent does not express a strong desire to "get away from" her / his home.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.5.3 Respondent's perception of home as "A place where I feel safe"

Definition

This measure is derived from responses to a questionnaire item which elicits the respondent's level of agreement with the statement "My house / flat is a place where I feel safe". The response options offered are the same as those for B.5.2 above. Responses were again dichotomised, contrasting the most favourable reply ("Strongly agree") against the remaining responses.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.5.4 Respondent's perception of home as "A place where I feel at home"

Definition

This measure is based on responses to a question which elicits the respondent's level of agreement with the statement "My house / flat is a place where I feel at home". The response options offered are the same as those for B.5.2 above. As with B.5.2 and B.5.3, responses were reduced to a binary quantity by contrasting the most favourable reply ("Strongly agree") against the remaining responses.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.5.5 Whether respondent would move home if able to do so

Definition

This measure is based on responses to the question "Would you move house if you were able to?". Only "Yes" and "No" responses are permitted. The analysis models the probability of a "No" response; that is, whether the respondent would not move house if able to do so.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.6 Drinking and smoking

B.6.1 Whether respondent has consumed an alcoholic drink in the last seven days

Definition

This measure is based on responses to the question "Have you had an alcoholic drink in the last 7 days?". Only "Yes" and "No" responses are permitted. The analysis models the probability of a "Yes" response.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.6.2 Whether respondent currently smokes cigarettes

Definition

This measure is derived from responses to a questionnaire item which elicits information on the respondent's use of cigarettes. Five response options are provided, namely "I have never tried smoking a cigarette, even a puff or two", "I have never really smoked cigarettes, just tried them once or twice", "I smoke cigarettes nowadays", "I do not smoke cigarettes at all nowadays, but I used to smoke regularly (at least one a day)" and "I do not smoke cigarettes at all nowadays, but I used to smoke occasionally (less than one a day)". For the analysis, the option identifying the respondent as a current smoker ("I smoke cigarettes nowadays") was contrasted against all other responses.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.7 Nature and extent of social contacts

B.7.1 Whether friends / relatives have been dissuaded from staying overnight due to poor housing conditions

Definition

This measure is derived from responses to the question "In the past four weeks have you put off friends / relatives coming to stay overnight because of poor housing conditions such as dampness or cold?". Response options are restricted to "Yes" and "No", the analysis modelling the probability of a "Yes" response.

Type of model / Additional predictors

Logistic regression / Predictor set B

B.7.2 Whether friends / relatives have been dissuaded from visiting due to poor housing conditions

Definition

This measure is derived from responses to a question similar to that used in B.7.1, the slightly-modified wording reading "In the past four weeks have you put off friends / relatives coming to see you because of poor housing conditions such as dampness or cold?". Response options are again restricted to "Yes" and "No", and the analysis predicts the probability of a "Yes" response.

Type of model / Additional predictors

Logistic regression / Predictor set B

B.7.3 Number of times respondent has gone out to visit family / friends in the past two weeks

Definition

This measure is based on responses to a question asking how many times the respondent has gone out to visit family or friends in the past two weeks. The response options provided are "Not at all", "Once or twice", "Three-six times" and "More than six times". For the analysis, the original responses were dichotomised such as to contrast "More than six times" against the remaining responses ( i.e. those indicating six or fewer contacts in the period).

Type of model / Additional predictors

Logistic regression / Predictor set A

B.7.4 Number of times respondent has been visited by family / friends at home in the past two weeks

Definition

This measure is based on responses to a question asking how many times the respondent has been visited by family or friends at home in the past two weeks. The response options provided are identical to those for B.7.3 and are reduced to a binary quantity in the same way ( i.e. contrasting "More than six times" against all other responses).

Type of model / Additional predictors

Logistic regression / Predictor set A

B.8 Perceived financial strain

B.8.1 Whether respondent reports any degree of financial difficulty

Definition

This measure is based on responses to the question "Taking everything together, which [phrase] best describes how you and your household are managing financially these days?". The response options provided are "Manage very well", "Manage quite well", "Get by alright", "Don't manage very well", "Have some financial difficulties" and "Are in deep financial trouble". For the analysis, original responses were reduced to a binary quantity by contrasting "Don't manage very well", "Have some financial difficulties" and "Are in deep financial trouble" against the remaining options. Reshaped in this way, the measure represents an indication of whether the respondent reports some degree of financial difficulty.

Type of model / Additional predictors

Logistic regression / Predictor set C

B.9 Specific symptoms and health conditions, and use of primary and secondary health services

B.9.1 Number of reported episodes of cold / flu symptoms in past 6 months

Definition

This measure is derived from responses to the question "How many times have you had cold or flu symptoms in the past six months?". Responses are collected as an absolute number of episodes.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.9.2 Whether respondent has ever been diagnosed with asthma

Definition

This measure is derived from responses to a question asking "Which, if any, of these health problems have you been diagnosed by a doctor as having?". Among the conditions specified is asthma; the other health problems featured are listed in B.9.3 to B.9.7 below. Because the question seeks to establish whether the respondent has ever been diagnosed with asthma (that is, has received a first diagnosis with the condition), analysis is restricted to those who remain 'at risk' of such a diagnosis at the study endpoint; that is, those who indicate a diagnosis of asthma at the initial interview are excluded.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.3 Whether respondent has ever been diagnosed with bronchitis etc.

Definition

As B.9.2, except that the condition specified is "Chest problems such as chronic bronchitis, pulmonary disease".

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.4 Whether respondent has ever been diagnosed with eczema

Definition

As B.9.2, except that the condition specified is "eczema".

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.5 Whether respondent has ever been diagnosed with nasal allergy

Definition

As B.9.2, except that the condition specified is "a nasal allergy such as hayfever".

Type of model

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.6 Whether respondent has ever been diagnosed with heart disease

Definition

As B.9.2, except that the condition specified is "heart disease".

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.7 Whether respondent has ever been diagnosed with circulatory problems

Definition

As B.9.2, except that the condition specified is "circulatory problems".

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.8 Number of reported attacks of asthma in the past 12 months

Definition

This measure is restricted to those respondents who report a clinical diagnosis of asthma at the initial interview (see B.9.2 above). Such respondents are presented with the question "How many attacks of asthma have you had in the past 12 months?". Responses are collected via a banded system offering the options "None", "One to three", " Four to ten" and "More than ten". For the analysis, the original replies were dichotomised at zero ( i.e. original response "None") versus any other response. The measure thus represents whether the respondent reports at least one asthma attack over the past year.

Type of model

Logistic regression

Additional predictors

For technical reasons, this analysis required to use a modified version of Predictor set A which excluded (a) household type and (b) the life event indicators.

B.9.9 Whether respondent has been woken by shortness of breath in the past 12 months

Definition

This measure is derived from responses to the question "Have you been woken by an attack of shortness of breath at any time in the past 12 months?". Response options are restricted to "Yes" and "No"; the analysis models the probability of a "Yes" response.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.10 Whether respondent has been woken by tightness in chest in the past 12 months

Definition

This measure is derived from responses to the question "Have you been woken up with a feeling of tightness in your chest at any time in the past 12 months?". Responses options are limited to "Yes" and "No", and the probability of the former is modelled by the analysis.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.11 Whether respondent has experienced wheezing in chest in the past 12 months

Definition

This measure is based on responses to the question "Have you had wheezing in your chest at any time in the past 12 months?". Only "Yes" and "No" responses are accepted, and the probability of the former is modelled by the analysis.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.12 Whether respondent experienced coughing or phlegm on most days

Definition

This measure is derived from responses to the question "Have you had coughing or phlegm on most days for a minimum of three months a year and for at least 2 successive years?". Responses are restricted to "Yes" and "No", and the analysis models the probability of a "Yes" response.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.13 Whether respondent suffers from at least one respiratory health problem

Definition

This measure is derived by determining whether the respondent provides a positive ("Yes") response to at least one of B.9.9 to B.9.11.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.14 Whether respondent has ever been diagnosed with high blood pressure

Definition

This measure is derived from responses to the question "Have you ever been diagnosed as having high blood pressure?". Response options are limited to "Yes" and "No". As with B.9.2 to B.9.7, respondents who report a diagnosis at the initial interview are excluded from analysis.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.15 Whether respondent has ever been advised to change diet or lifestyle due to high blood pressure

Definition

This measure is based on responses to the question "Have you ever been advised by a health professional to change your diet or lifestyle to reduce your blood pressure or avoid having high blood pressure?". Only "Yes" and "No" responses are offered. Respondents who report "Yes" at the initial interview are again excluded. The analysis models the probability of a "Yes" response.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.16 Whether respondent is currently taking action in relation to diet or lifestyle due to high blood pressure

Definition

This measure is based on responses to the question "Are you currently taking any action in relation to your diet or lifestyle because of concerns about your blood pressure?". Only "Yes" and "No" responses are provided, and the analysis models the probability of the former.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.17 Whether respondent is currently suffering from high blood pressure

Definition

This measure is derived from responses to the question "As far as you are aware, do you have high blood pressure at the present time?". Responses are restricted to "Yes" and "No", and the analysis models the probability of a "Yes" reply.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.18 Number of GP / nurse encounters in past year

Definition

This measure is derived from responses to the question "Over the past year how often have you seen or spoken to a GP or nurse either at their practice or at home about yourself?". Responses are collected as an absolute number of encounters.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.9.19 Number of hospital outpatient or day bed visits in past year

Definition

This measure is derived from responses to the question "Over the past year how often have you visited a hospital outpatient clinic or day beds?". While responses are collected as an absolute number of visits, the original values were (for technical reasons) dichotomised for analysis at zero versus one or more visits. Reshaped in this way, the measure represents whether the respondent has experienced at least one hospital outpatient or day bed visit in the past year.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.20 Number of overnight hospital stays in past year

Definition

This measure is based on responses to the question "Over the past year how often have you had an overnight stay in hospital?". As with B.9.19, the original responses were reduced to a binary quantity - zero versus one or more stays - for analysis.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.9.21 Number of Accident and Emergency attendances in past year

Definition

This measure is derived from responses to the question "Over the past year how often have you visited a Casualty or Accident and Emergency Department for treatment for yourself?". As with B.9.19, the original responses were dichotomised at zero versus one or more stays for analysis.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.10 Self-reported health-related quality of life

B.10.1 SF-36 Physical Functioning scale

Definition

This measure is the respondent's score on the Physical Functioning scale of the SF-36 Version 2 Health Survey. The SF-36 is a widely-used and well-validated questionnaire which elicits information on various dimensions of health and well-being. Full details of the SF-36 are given in [ref. 3].

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.10.2 SF-36 Role-Physical scale

Definition

This measure is the respondent's score on the Role-Physical scale of the SF-36 Version 2.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.10.3 SF-36 Bodily Pain scale

Definition

This measure is the respondent's score on the Bodily Pain scale of the SF-36 Version 2.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.10.4 SF-36 General Health scale

Definition

This measure is the respondent's score on the General Health scale of the SF-36 Version 2.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.10.5 SF-36 Vitality scale

Definition

This measure is the respondent's score on the Vitality scale of the SF-36 Version 2.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.10.6 SF-36 Social Functioning scale

Definition

This measure is the respondent's score on the Social Functioning scale of the SF-36 Version 2.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.10.7 SF-36 Role-Emotional scale

Definition

This measure is the respondent's score on the Role-Emotional scale of the SF-36 Version 2.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.10.8 SF-36 Mental Health scale

Definition

This measure is the respondent's score on the Mental Health scale of the SF-36 Version 2.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.10.9 SF-36 Health Transition item

Definition

This measure is the respondent's score on the Health Transition item of the SF-36 Version 2.

Type of model / Additional predictors

Analysis of covariance / Predictor set A

B.11 Long-standing illness or disability

B.11.1 Whether respondent suffers from long-standing illness or disability which limits daily activities / work

Definition

This measure is derived from responses to the question "Do you have any long-standing illness, health problem or disability that limits your daily activities or the kind of work you can do? By disability as opposed to ill-health, I mean a physical or mental impairment, which has a substantial and long-term adverse effect on your ability to carry out normal day-to-day activities.". Response options are limited to "Yes" and "No", and the analysis models the probability of the former.

Type of model / Additional predictors

Logistic regression / Predictor set A

B.12 Use of medications

B.12.1 Whether respondent is currently taking prescribed medications

Definition

This measure is derived from responses to the question "Are you currently taking any medicines prescribed to you by a doctor, including inhalers?". Responses take the form of "Yes" or "No".

Type of model / Additional predictors

Logistic regression / Predictor set A

B.12.2 Whether respondent is currently taking 'over the counter' medications

Definition

This measure is based on responses to the question "Are you currently taking any medicines that you bought yourself, without a prescription?". Responses are restricted to "Yes" and "No".

Type of model / Additional predictors

Logistic regression / Predictor set A

B.13 Additional predictors included in analyses

Predictor set A:

a) A binary indicator denoting whether the respondent is a central heating recipient or a comparison group respondent (coding: 0 = comparison, 1 = recipient). Estimated values of this parameter are the key results obtained from this study.

b) The age of the respondent ( i.e. the household member who actually gave the initial interview, and completed the follow-up postal questionnaire), in years.

c) The gender of the respondent (coding: 0 = male, 1 = female)

d) Socioeconomic group (coded as a set of three binary indicators, denoting membership [0 = NO, 1 = YES] of classes AB; C1; and C2. When all three indicators hold zero values, membership of the reference class DE is indicated).

e) Household type (coded as a set of six binary indicators, denoting classification of household [0 = NO, 1 = YES] as single adult; single parent; couple without children; couple with children; pensioner couple; and multiple adult. The reference class of single pensioner is represented by zero values for all six indicators).

f) A simplified representation of housing tenure (coded 0 = owner-occupier, 1 = renter or other tenure type)

g) A group of five binary indicators representing whether the respondent has experienced specific life events during the year prior to the final interview. The events included were (i) serious illness, requiring hospitalisation, experienced by the respondent or a relative / close friend; (ii) divorce, separation or break-up of an intimate relationship; (iii) bereavement (death of a relative / close friend); (iv) a period of unemployment, of at least one month's duration, experienced personally by the respondent; and (v) a period of unemployment, of at least one month's duration, experienced by another wage earner in the respondent's household. These events are represented in the model by binary indicators coded 0 = NO (event not experienced), 1 = YES (event experienced).

h) A representation of change in smoking exposure. This is implemented as three binary indicators, as follows. Element (i) indicates whether the respondent actively smokes at the study endpoint (coded 0 = NO, 1 = YES). Element (ii) indicates whether the respondent is exposed to passive smoking in the home at the study endpoint (coded 0 = NO, 1 = YES). Element (iii) indicates whether the respondent's exposure to smoking has changed between the first and final data collection waves (coded 0 = NO, 1 = YES).

Predictor set B:

As Predictor Set A, but with the representation of social group and of smoking exposure excluded.

Predictor set C:

As Predictor Set A, but with the representation of smoking exposure excluded.