Smoking

At baseline, the majority (88.1%) of participants smoked cigarettes or tobacco (not including cannabis or other drugs). A further 4.4% used to smoke but had stopped. In comparison, just 13.9% of adults in England in 2019 were smokers.¹²⁷ It is difficult to collect precise information about smoking habits, although 21.1% of participants reported having more than or equal to 20 cigarettes or ‘roll-ups’ daily and so were classified as ‘heavy smokers’. Participants of Mobile Teams were least likely, and those of Specialist GPs most likely, to report current smoking; the findings were statistically significant (Table 28). Of those who smoked at baseline, 34.8% had received advice or help in the preceding 4 months with reducing or stopping smoking from CSS doctors or nurses, including 15 participants who had been given nicotine patches. Usual Care GP participants were more likely to have received help, although the findings were not statistically significant.

TABLE 28

Smoking and nutrition at baseline by Health Service Model

Thirteen participants who were smokers at baseline had stopped smoking by 8 months: four from Mobile Teams and three from each of the other models. Nineteen people who smoked, but not heavily, at baseline reported being heavy smokers at 8 months (four from Dedicated Centres, eight from Mobile Teams, five from Specialist GPs and two from Usual Care GPs). This may reflect an increase in their tobacco intake, or they were more candid over time about smoking habits. In contrast, 21 people who were heavy smokers at baseline smoked less than 20 per day at 8 months. Of those who did not smoke at baseline, none had started or resumed smoking at 8 months.

Nutrition

At baseline, the median number of days per week that participants had a meal containing protein, such as meat, fish or a vegetarian equivalent, was five. This varied greatly: 48.5% had such a meal 6 or 7 days a week, and 20.5% had such a meal once a week or less (see Table 28). Likewise, 25.1% had a portion of fruit or vegetables once a week or less. Just over one-third of participants cooked for themselves, 23.5% had food provided at their hostel and 45.7% had food at day centres. Other sources of food included soup runs and handouts on the streets (17.7%), takeaway food (13.9%) and meals cooked by relatives or friends (10.8%). Overall, 53.6% reported difficulty getting meals or eating healthily, mainly because of insufficient income, poor appetite or problems with cooking. Three-tenths had used a food bank at least once in the preceding 4 months. Participants in their own tenancy, followed by those sleeping rough, were more likely to report poor food intake. Among the former, 34.6% had a meal once a week or less, 63% described difficulty getting meals or eating healthily, and several relied on day centres for food. Participants who were sleeping rough mainly obtained food at day centres and street handouts.

There were no statistically significant differences by Health Service Model in nutrition habits among participants, although a slightly higher percentage of Usual Care GP participants reported difficulty getting meals or eating healthily. Nearly three-tenths of participants (28.5%) had received help in the preceding 4 months regarding nutrition from a CSS doctor or nurse, including advice about weight and food intake, vouchers for food banks or a prescription for nutritional supplements. Dedicated Centres and Specialist GPs were most likely, and Usual Care GPs least likely, to have provided help; the findings were statistically significant (see Table 28). Hence, Usual Care GP participants were most likely to describe nutritional problems and least likely to have received advice from the CSS.

By 8 months, the median number of days per week that participants had a meal had increased to seven, with 17% stating once a week or less. There were no statistically significant differences by Health Service Model in nutrition habits at 8 months, although slightly fewer (40.8%) described problems obtaining food or eating healthily.

Health status

The SF-8 is a short, self-administered instrument that produces a health-related quality-of-life profile. It measures eight ordinal items over the preceding 4 weeks: general health, physical functioning, ability to carry out physical activities, bodily pain, vitality, social functioning, mental health and emotional roles. Scale means, based on the same standard metrics as for the Short Form 36 Health Survey, are assigned to each response, creating a continuous outcome for each of the eight items. Regression coefficient weights are assigned to each item to produce a PCS and a MCS, both of which are continuous variables and have a mean of 50 (SD 10) in the general population of the USA.¹²⁸ Higher scores represent more favourable health states, and lower scores indicate worse health states.

The SF-8 was self-completed by the HEARTH study participants at baseline and at 8 months. At baseline, the mean PCS was 39.8 and the mean MCS was 35.4, with no statistically significant difference by Health Service Model. There was a slight increase at 8 months in the mean PCS and MCS across all models, but the findings were not statistically significant. When changes in scores are examined for participants who completed the SF-8 at both baseline and 8 months, there were no statistically significant differences by Health Service Model, although Mobile Team and Usual Care GP participants experienced a negative change in mean for the SF-8 MCS (Table 29).

TABLE 29

The SF-8 and SWEMWBS scores: (1) at baseline and 8 months, and (2) change in scores during this period by Health Service Model

The average scores in the general population of the USA for the PCS and the MCS are 50.¹²⁸ Of the HEARTH study participants, at baseline, 81.9% scored less than this for the PCS and 83.6% scored less than this for the MCS. No data could be found of SF-8 use in England among either the general population or people who were homeless to compare scores. However, a recent study in Belfast used the SF-8 to assess the mental health of people in different deprivation areas, drawing on the 2010 Northern Ireland Multiple Deprivation Measure.¹²⁹ The MCS score for those in the most deprived area (n = 215) was 45.6, which is considerably higher than that of the HEARTH study participants. Likewise, during a survey in Alameda County, California, 292 people who had been homeless for 12 months or more in the previous 3 years (i.e. chronically homeless) completed the SF-8. Their scores were slightly higher (PCS 42.1, MCS 39.6) than those of the HEARTH study participants.¹³⁰

Mental well-being

The SWEMWBS was developed in 2009 as a shorter version of the Warwick–Edinburgh Mental Wellbeing Scale. It is a validated instrument for use by the general population, and has seven items relating to psychological functioning, such as feelings of usefulness and optimism, thinking clearly and dealing with problems.¹³¹ Each item allows five responses, from ‘none of the time’ to ‘all of the time’ (scored 1–5), and people rate their experiences over the preceding 2 weeks. Scores range from 7 to 35, with higher scores indicating greater positive mental well-being. The total raw scores from the SWEMWBS are transformed into metric scores using a conversion table.

The SWEMWBS was self-completed by the HEARTH study participants at baseline and 8 months. At baseline, the mean score was 18.5, with no statistically significant difference by Health Service Model. There was a slight increase in scores for all models at 8 months, but the findings were not statistically significant. Likewise, there were no statistically significant differences by Health Service Model in changes in scores between baseline and 8 months when the scores of the 248 participants who completed the SWEMWBS at both baseline and 8 months are examined (see Table 29).

When the scores are compared with responses in the Health Survey for England 2010–13 of 27,169 people aged 16 years and older,¹³¹ the mental well-being of the HEARTH study participants was considerably lower than that of the general population. In the HEARTH study, the mean baseline scores were 18.4 for men and 18.7 for women. In contrast, the mean scores in the Health Survey for England were 23.7 for men and 23.2 for women.¹³¹

Regression modelling of Short Form 8 Health Survey and Short Warwick–Edinburgh Mental Wellbeing Scale scores

Dichotomous and continuous predictors, similar to those applied to the modelling of the Primary Outcome Score, were used to isolate the effect of the four Health Service Models in the modelling of two outcomes for each of the following: the SF-8 PCS, the SF-8 MCS and the SWEMWBS. The first concerns baseline scores, and the second concerns changes in scores from baseline to 8 months.

Baseline scores

For modelling the baseline scores, an initial backward stepwise linear regression was run with each dependent variable at baseline and the variables in Appendix 3, Tables 48 and 49, as predictors. Statistically significant and near statistically significant (p < 0.15) predictors were used in a second backward stepwise linear regression. A final model was run using only the statistically significant predictors resulting from this regression to maximise the number of observations used.

The salient fitted parameters, namely those for Health Service Model (whether or not statistically significant) and for other statistically significant predictors for the SF-8 and the SWEMWBS outcomes, are shown in Table 30. No statistically significant beneficial effect is seen when comparing Dedicated Centre, Mobile Team and Specialist GP models with the Usual Care GP model at baseline. Across the three secondary outcomes, various other predictors show statistical significance (albeit never more than twice), reflecting the importance of taking into account many facets of the target population before assessing the efficacy of the four Health Service Models. For example, physical health problems and an increasing number of GP or nurse consultations were negatively associated with the SF-8 PCS, which is understandable given that it concerns physical functioning. Increasing age and drug use also had similar effects, both of which are likely to affect physical health.

TABLE 30

Baseline secondary outcome regressions for SF-8 and SWEMWBS: resulting models

In terms of the SF-8 MCS and the SWEMWBS scores which concern mental health and psychological well-being, self-reported depression had a significant negative effect, which is to be expected. Likewise, there was a negative association with both outcomes among participants who injected drugs. In relation to the SWEMWBS, statistically significant negative effects were also linked to heavy drinking and increasing length of time homeless, both of which are likely to have an adverse impact on morale and well-being.

Changes after 8 months

When changes in the SF-8 PCS, the SF-8 MCS and the SWEMWBS scores are examined from baseline to 8 months, one or more changes was observed for 253 participants (69.7%), with 244 (67.2%) participants having changes for all three outcomes. Dichotomous and continuous predictors were used

to isolate the effect of the four Health Service Models in the modelling of changes in SF-8 PCS, SF-8 MCS and SWEMWBS scores (see Appendix 3, Tables 50 and 51). Variables relating to alcohol and drug use, smoking, income, involvement in education or employment, and informal support relate to circumstances at 8 months. Baseline variables for the principal predictors concerning physical and mental health were retained, as the corresponding variables at 8 months confound with the SF-8 PCS and SF-8 MCS outcomes at 8 months.

An initial backward stepwise linear regression was run, and statistically significant and near statistically significant (p < 0.15) predictors were used in a second backward stepwise linear regression. A final model was run using only the statistically significant predictors resulting from this regression to maximise the number of observations used. The salient fitted parameters, namely those for Health Service Model (whether or not statistically significant) and for other statistically significant predictors for the SF-8 and the SWEMWBS outcomes, are shown in Table 31.

TABLE 31

Secondary outcome regressions for the SF-8 and the SWEMWBS: resulting models for change from baseline to 8 months

When comparing changes from baseline to 8 months, Dedicated Centres and Specialist GPs both elicit a comparative improvement of around 4 points for the SF-8 PCS, and there is also an improvement in the SWEMWBS scores for Specialist GPs. These findings are statistically significant. Increasing age, heavy drinking, depression and an increasing number of GP or nurse consultations at the CSS were associated with a negative effect on change in SF-8 PCS. Increasing age and heavy drinking are likely to contribute to physical health problems, which, in turn, are likely to lead to consultations with primary health care professionals. The large negative depression coefficients (−4.380 and −2.462) illustrate the ongoing impact of depression on physical and mental well-being.

There was a negative association between change in the SF-8 MCS and the percentage of time living in staffed accommodation during the study period. The likely explanation is that participants with mental health problems spent more time in staffed accommodation during the study than those without mental health problems (mean percentage of time 49.98 and 27.27, respectively). As with the baseline regressions, self-reported depression and injection of drugs were negatively associated with beneficial change in SWEMWBS scores, whereas there was a strong positive effect from involvement in education, training or employment. The latter is likely to be associated with increased motivation and morale, which, in turn, will influence psychological functioning.

Summary

This chapter has examined the impact of the various Health Service Models on health status and well-being over time. A very high percentage of participants were smokers, and many had poor nutrition. There were slight improvements over time in their smoking and eating habits, but no statistically significant differences by Health Service Model in the provision of help for smoking. However, Dedicated Centres and Specialist GPs were most likely, and Usual Care GPs least likely, to have provided help with nutrition: a statistically significant disparity.

The SF-8 PCS, the SF-8 MCS and the SWEMWBS scores of the HEARTH study participants indicate low levels of physical and psychological functioning, compared with the general population and other marginalised groups. There were no significant differences in their scores at baseline by Health Service Model, and most models experienced a slight improvement in scores over time. Further interrogation of the data through regression modelling, however, revealed a positive improvement in changes in SF-8 PCS from baseline to 8 months by Dedicated Centres and Specialist GPs, and a positive change in SWEMWBS scores for Specialist GPs. However, several participant characteristics had strong significant effects on scores at baseline, and on changes in scores from baseline to 8 months. These mainly involved negative effects associated with older age, physical health problems, depression, heavy drinking, drug use and injection of drugs. A strong positive effect concerning involvement in education, training or employment was associated with change in the SWEMWBS scores from baseline to 8 months. The next chapter focuses on oral health and participants’ use of dental services.