Can alcohol consumption in Germany be reduced by alcohol screening, brief intervention and referral to treatment in primary health care? Results of a simulation study
Summary (4 min read)
Introduction
- Globally, alcohol use is a major risk factor for the burden of mortality and disease [1, 2] .
- Various initiatives have been set by the World Health Organization (WHO) and the United Nations has set objectives to reduce this burden [3] .
- The only application of a simulation model to quantify the effects of SBI in Germany known to the authors was carried out as part of a report issued by the Organization for Economic Co-operation and Development [17] .
- In fact, none of the 22 studies identified in the 2014 review disclosed their simulation programmes [16] .
Attenuation of effects
- The attenuation of intervention effects, assuming that BI effects to remain stable for a period of four years and to attenuate thereafter and reach 0 after ten years (linearly imputed for all years five to ten) according to [25] and [26].
- Lastly, sex, age, and education stratified population data for all years was obtained by combining population data from UN (providing data for all age groups and years, but not by education) and EUROSTAT (providing education breakup for all years, but only for age groups up to 74 years).
- As a result, the authors obtained a complete set of prevalence estimates for lifetime abstinence, former drinking, current drinking, and HED, as well as average drinking levels per drinker -for all years 2009 to 2018 and by sex, age, and education.
- These data served as input data for the next step.
Step 2: Simulating baseline alcohol consumption in the population
- For the year 2009, a population sample of 100 persons was drawn, stratified by sex, age, and educational level.
- Using binomial distributions, the drinking status prevalence estimates described in step 1 were used to determine the drinking status (either lifetime abstainer, former drinker, or current drinker) for each person.
- Second, the authors determined daily drinking levels (in grams pure alcohol per day) for each current drinker, which was drawn from a gamma distribution, which has been shown to approximate alcohol use self-reports from surveys [30, 31] .
- Lastly, HED status was determined for each current drinker based on the data from step 1, again using binomial distributions.
Step 3: Applying effects of SBIRT
- In order to apply the effects of SBIRT, the following four conditions had to be fulfilled: Persons had to: attend PHC; be screened for alcohol use; drink riskily; and receive a BI (for medium risk drinkers) or RT (for high-risk drinkers).
- In the following, each consecutive conditional step is described in detail.
- Based on these prevalence estimates, binomial distributions were used to determine whether a person had at least one PHC visit in the current year.
- In addition to manipulating the drinking levels for positively screened persons receiving BI or RT, the HED status was also changed based on effect sizes from the same metaanalysis (see Table 1 for effect sizes).
Step 4: Accounting for secular changes and attenuating intervention effects
- Step 3, i.e., the application of SBIRT effects, was repeated for each year.
- Prior to applying the effects in each year, the authors accounted for a) attenuating intervention effects from previous years, and b) secular changes in APC and drinking status prevalence.
- First, the authors assumed that intervention effects on drinking levels remained stable for a period of four years (according to [25] ), attenuate thereafter, and nullify after ten years (according to [26] , linearly imputed for all years five to ten).
- For drinkers giving up HED following BI or RT, the authors assumed that the chance to re-engage in HED was 50% chance starting from the second year post intervention.
- Second, the authors corrected drinking levels among drinkers and drinking status to match the observed trajectories in the input data.
Sensitivity analysis
- Keeping all other parameters constant, the authors tested the impact of the attenuation of intervention effects over time.
- In an additional sensitivity analysis, the authors assumed that any intervention effect diminishes three years post intervention.
Reporting the simulation findings
- The outcomes of interest were drinking levels and prevalence of HED, which are the two alcohol exposure variables known to be most impacted by BI and RT delivered in PHC settings.
- All findings are reported against the baseline scenario for the final year 2018, thus describing the cumulative effects over a ten-year period.
- The complete R code including input data is appended to this paper to allow for complete reproducibility and adjustment of parameters to other settings (see S1 file).
- As input data for the simulation, the authors obtained aggregated and fully anomized secondary data from previous surveys, which have undergone formal ethical reviews (for details, see [21, 23] ).
Alcohol exposure in Germany in the baseline scenario
- Between 2009 and 2018, alcohol consumption hardly changed in Germany.
- The green line was taken from the scenario without any screening activity, thus, the drinking level follows the observed trend of drinking levels in this population, as specified in the input data.
- The blue line was taken from the baseline scenario, thus, at a screening rate of 2.9%.
Estimated coverage of SBIRT
- Across the ten-year period, every 40 th adult (2.4%, 95% CI: 1.8% to 3.1%) was estimated to have benefitted from an intervention, i.e., BI for medium risk and RT for high-risk drinkers, following alcohol screening in PHC in the baseline scenario.
- The screening and intervention rates achieved by end of the ten-year period are illustrated in Fig 3 .
Impact on APC and HED
- In Tables 2 and 3 , the simulation results are presented for two outcomes of interest, the mean daily drinking levels and the prevalence of HED in the adult population.
- Presented are relative changes of mean drinking levels to the as-is-scenario, with a screening rate of 2.9%.
- Bold results indicate confidence intervals not overlapping with 0, indicating significant differences to the baseline scenario.
- Significant reductions of drinking levels could have been achieved for five out of twelve subgroups if one out of two PHC patients were screened for their alcohol use: 15 to 34 year olds with low and middle education levels, 35 to 49 year olds with high education levels, and 65 to 99 year olds with low and high education levels.
Sensitivity analyses
- In sensitivity analyses, the authors modeled the impact of SBIRT under the more conservative assumption according to which the intervention effects would completely diminish three years post intervention, as compared to the slower attenuation beginning only five years post intervention as implemented in the main analyses.
- As illustrated in Fig 6 , the scenarios of 0 to 25% screening coverage are robust to the underlying assumption, i.e., changing the assumption would have no significant impact on the estimated drinking levels.
Limitations
- Before further discussing the findings of this study, the authors need to highlight several limitations.
- First and foremost, as with any simulation study, the authors rely on assumptions that may not hold true.
- The authors have attempted to be transparent with all assumptions, reporting all parameters used in the simulation, and performing rigorous sensitivity analyses to test one key assumption.
- Second, the authors tried to rely mostly on local data, except for effect sizes and attenuation parameters.
- BI reception was more often reported by socioeconomically disadvantaged drinkers in England [33] , and if similar patterns were present in Germany, this would change the simulation findings accordingly.
Comparison with other simulation studies
- Several other simulations have quantified the effects of scaling up SBIRT (for an overview, see [16] ), however, the authors are only aware of one application for Germany.
- In a 2015 report, the Organization for Economic Co-operation and Development, the effects of BI were estimated for a screening coverage of 40% and a 30% intervention probability for positive screened patients, with the effects waning within 12 months after receiving the intervention [17].
- In their simulation, which was performed for a 40-year time period, the prevalence of hazardous/harmful drinking could be reduced by 5%, while their results suggest a reduction in per capita consumption by 12% in the most comparable scenario of a 50% screening rate over a period of 10 years.
- The proposed simulation methodology can be extended as well to address health economic issues, including costeffectiveness analyses.
- In the UK, for example, an assessment of 1.8 million patient records in 2018 found that 48.8% of adult patients had a measure of alcohol consumption recorded during the previous five years [42] .
Implications for alcohol policy in Germany
- The simulation results suggest that the current coverage of alcohol screening hardly matters for population alcohol exposure in Germany.
- The authors show that the large-scale delivery of SBIRT in German PHC settings could be a viable measure to accelerate the ongoing trend.
- In a comparative survey, nearly half of German general practitioners did not consider alcohol as an important risk factor for hypertension, in contrast to a share of 15% among respondents from France, Italy, Spain, and the UK [45] (for evidence on alcohol use and hypertension, see [46] ).
- While further efforts are needed to increase SBIRT delivery in German PHC settings in the long run, e.g., by financial reimbursement of alcohol management activities [50] , alternatives may be required to reduce alcohol consumption and attributable burden in the short-term.
- Evidence from Lithuania and Great Britain, for example, demonstrates the impact that policies targeting alcohol prices can have in reducing consumption and harm [54] [55] [56] [57] .
Did you find this useful? Give us your feedback
Citations
5 citations
2 citations
1 citations
References
1,226 citations
530 citations
469 citations
436 citations
272 citations
Related Papers (5)
Frequently Asked Questions (18)
Q2. How many drinkers were present in Germany in 2018?
In 2018, 230 73.8% of drinkers were estimated to have low risk drinking levels, while medium and high-risk 231 drinking was present among 11.9% and 14.3% of drinkers.
Q3. How high is the coverage of alcohol measurement in integrated health care systems?
in integrated health-care systems where alcohol measurement is mandated and built into the electronic medical record system, as it is in the US Veterans Health Administration system, coverage can be as high as 93% [43].
Q4. How many adults had a measure of alcohol consumption recorded in 2018?
In the UK, for example, an assessment of 1.8 million patient records in 2018 found that 48.8% of adult patients had a measure of alcohol consumption recorded during the previous five years [42].
Q5. How can the authors use the model to perform health economic studies?
If combined with health outcomes, e.g. by using the open-access programme InterMAHP [38], their simulation model can be readily adapted to perform health economic studies, such as cost-effectiveness analyses.
Q6. How many people would have received a subsequent intervention by the end of the ten-year?
If every second PHC 46 patient would have been screened for alcohol use, 21% of adult residents in Germany would 47 have received BI or RT by the end of the ten-year simulation period.
Q7. How many people were screened for alcohol use in Germany in 2016?
70 In Germany, application of SBIRT is recommended by the ‘Guidelines on Screening, 71 Diagnosis and Treatment of Alcohol Use Disorders’ [11], however, survey data from the federal 72 state of Bremen suggest that only 2.9% of patients were screened by their primary health care 73 (PHC) providers in 2016 [12].
Q8. How many years of intervention would the effect of SBIRT be reduced?
In sensitivity analyses, the authors modeled the impact of SBIRT under the more conservativeassumption according to which the intervention effects would completely diminish three years post intervention, as compared to the slower attenuation beginning only five years post intervention as implemented in the main analyses.
Q9. How many men and the youngest age groups could be screened for alcohol use?
If alcohol use was assessed in every fourth patient, reductions in drinking levels among 280 men and the youngest age groups could be achieved.
Q10. How could Germany achieve a large-scale implementation of SBIRT?
A large-scale implementation of SBIRT in Germany could only be achieved in the more distant future, thus, other alcohol policy options should be considered as well to achieve short-term reductions in alcohol consumption.
Q11. What was the only application of a simulation model to quantify the effects of SBI in Germany?
The only 80 application of a simulation model to quantify the effects of SBI in Germany known to the authors 81 was carried out as part of a report issued by the Organization for Economic Co-operation and 82 Development [17].
Q12. How did the authors determine drinking levels for each current drinker?
the authors determined daily drinking levels (in 152 grams pure alcohol per day) for each current drinker, which was drawn from a gamma 153 distribution, which has been shown to approximate alcohol use self-reports from surveys [30, 154 31].
Q13. How many people were screened for alcohol in the baseline scenario?
258 Across the ten-year period, every 40th adult (2.4%, 95% CI: 1.8% to 3.1%) was estimated 259 to have benefitted from an intervention, i.e., BI for medium risk and RT for high-risk drinkers, 260 following alcohol screening in PHC in the baseline scenario.
Q14. How many people received a subsequent intervention between 2009 and 2018?
44 In the baseline scenario of 2.9% screening coverage, 2.4% of the adult German 45 population received a subsequent intervention between 2009 and 2018.
Q15. How many people were screened for alcohol in Germany in the period 2009 to 2018?
Their findings suggest that screening up to one tenth of patients per year would not have significantly changed how alcohol consumption has developed in Germany in this time period.
Q16. How did the authors calculate the chance to re-engage in HED?
For drinkers giving up HED 194 following BI or RT, the authors assumed that the chance to re-engage in HED was 50% chance starting 195 from the second year post intervention.
Q17. How can the authors reduce alcohol consumption in Germany?
While further efforts are needed to increase SBIRT delivery in German PHC settings inthe long run, e.g., by financial reimbursement of alcohol management activities [50], alternatives may be required to reduce alcohol consumption and attributable burden in the short-term.
Q18. How much would the APC be at the end of the ten-year period?
in the scenarios of 50% and 75% screening coverage, the more conservative assumption would result in APC at the end of the ten-year simulation period to be, respectively, 6.1% (1.5 to 11.4%) and 6.9% (2.5 to 10.4%) higher.