Life’s Essential 8 in relation to self-rated health and health-related quality of life in a large population-based sample: the SCAPIS project

LE8 was defined based on the AHA criteria and incorporates 4 health behaviors: diet, physical activity, nicotine exposure, and sleep health, and 4 health factors measurements: BMI, non-HDL cholesterol, fasting blood glucose/glycosylated hemoglobin, and blood pressure [5]. Details about measurement and calculation of health behaviors and health factors in LE8 and LS7 have been published elsewhere [20].

In brief, regarding health behaviors, dietary habits were evaluated using the web-based questionnaire (MiniMeal-Q) and the scores were adapted from the Mediterranean Eating Pattern for Americans (MEPA) [26]. Physical activity was measured over a 7 days period with three different tri-axial accelerometers: Actigraph GT3X + , wGT3X +,  and wGT3X-BT (ActiGraph LCC, Pensacola, FL, USA) [27], considering ≥ 2690 counts per minute as moderate-vigorous intensity physical activity [28]. Nicotine exposure and sleep health were assessed with self-administered questionnaires. For health factors, standardized laboratory and clinical procedures were used, and medication consumption was collected from the Swedish Prescribed Drug Register.

To calculate an overall LE8 score, all 8 components within LE8 were scored from 0 to 100 (0, the worst health; 100, the best health). Following the AHA recommendations, an overall LE8 score was calculated as the unweighted average of all present components (range: 0–100). A minimum of 7 components was required for computing the overall LE8 score. In addition, two different scores, ranging from 0 to 100, were calculated for LE8 behaviors and LE8 factors as the unweighted average of all present components in behaviors and factors, respectively. In both cases, a minimum of 3 reported components was required for computing behaviors and factors scores.

LS7 was defined in accordance with the AHA criteria and incorporates 4 health behaviors: diet, physical activity, BMI, and smoking status, and 3 health factors: total cholesterol, blood glucose, and blood pressure [4]. In SCAPIS, with the exception of sleep health, the measurements, techniques, and questionnaires to obtain the LS7 score were similar to those in LE8, though the calculation of scores differed [20]. In LS7, as the AHA recommended, dietary habits were consistent with the Dietary Approaches to Stop Hypertension (DASH) eating plan [4].

To compute the LS7 score, all 7 components were required for computation, and two scoring algorithms were created. In the LS7 (0–7) score, the total score is equal to the number of components at the ideal level, with 0 representing the worst health and 7 representing the best health status. The LS7 (0–14) score is calculated as the sum of components scored as “poor = 0,” “intermediate = 1,” and “ideal = 2 points,” yielding a total score from 0 to 14, with 0 representing the worst health and 14 representing the best health status. In addition, two different scores were calculated for LS7 behaviors and LS7 factors as the sum of all present components in behaviors and factors, respectively. All 4 behaviors and all 3 factors were necessary to compute behaviors and factors scores, respectively.

In SCAPIS, from the Swedish SF-12.V1 questionnaire, which is an adaptation of the Swedish SF-36 questionnaire [29‐31], three different scores were derived to operationalize SRH and HRQoL. The construct SRH was measured with the 1-item question “In general, would you say your health is …?”, which was scored on a 5-point Likert scale as: poor, fair, good, very good, and excellent. Despite its simplicity, this 1-item self-reported question (hereafter referred to as “SRH”) has proved good validity to measure “general health,” adequately integrating the physical, psychological, and social dimensions in health [6]. In accordance with previous publications [32‐34], the SRH score was dichotomized into poor (poor and fair) and good (good, very good or excellent).

By contrast, HRQoL is a multidimensional index referring to the physical, psychological, and social domains of health and well-being [7]. The Swedish SF-12.V1 questionnaire aggregates the eight subscales of SF-36 allowing for the calculation of two different summary scores: “mental-HRQoL” and “physical-HRQoL” [35]. Both scores were computed following a modified protocol based on Ware’s framework [35], utilizing Farivar’s weights [36] instead of Ware’s original weights. We did not use Ware’s orthogonally rotated weights, as they force mental and physical scores to be uncorrelated, despite evidence of their non-independence [37]. Thus, the validity and interpretation of the orthogonal approach have been criticized [36‐39]. Instead, we used weights derived from oblique (correlated) factor analysis as Farivar et al. proposed [36]. In brief, we initially reversed the SCAPIS SF-12.V1 scores to ensure higher values consistently indicated better health. We then utilized dummy variables to categorize item responses, and later multiplied scores in dummy variables by suggested weights, incorporating a constant.

In the Swedish SF-12.V1 questionnaire, questions 9 to 12 (9, “Feelings last 4 weeks: felt calm and peaceful”; 10, “Feelings last 4 weeks: had a lot of energy”; 11, “Feelings last 4 weeks: felt downhearted and blue”; and 12, “Health limitations, last 4 weeks: interference with social activities”) only had five response options (0, “All of the time”; 1, “Most of the time”; 2, “Some of the time”; 3, “A little of the time”; 4, None of the time”) instead of six as in the original Ware’s manual [35]. In this manual, 5 weights (response options minus 1) were considered to score the 6-items 9–12 questions. In our work, we assigned Farivar’s weights for first (1) and last category (5) to our first (1) and last category (4), and then we linearly distributed the difference between first and last categories into the second and third categories [36]. The cutoffs for poor mental-HRQoL and poor physical-HRQoL were defined as the sample mean minus 1 standard deviation.

We conducted a complete case analysis, excluding participants with incomplete data on exposures, outcomes, and main covariates in the main model (i.e., Model 2). Descriptive statistics are presented as means and standard deviations for the continuous variables or frequencies and percentages for the categorical variables.

First, we examined the distribution of the levels of self-rated health along Life´s Essential 8 categories (Fig. 1). Second, we examined associations of LE8 and LS7 with SRH through binary logistic regression models since SRH is typically characterized on a dichotomous scale (i.e., excellent/very good/good vs fair/poor) (Fig. 2). To allow for potential non-linear associations, we modeled the associations using restricted cubic splines with 4 knots [40] located at percentiles 5th, 35th, 65th, and 95th across LE8 and LS7, respectively. Reference values in LE8, LE8 behaviors, and LE8 factors were settled at 80 points, which aligns with the cut-off proposed by the AHA to define high cardiovascular health [5], see corresponding percentiles in Fig. 2. For LS7, to maximize the distributional comparability to the 80-point cut-off in LE8, the references were settled at 4 and 10 points for LS7 (0–7) and LS7 (0–14), respectively. Similarly, reference for LS7 behaviors and factors were settled at the point that render the most similar distribution to the global LS7 score, see corresponding percentiles in Supplementary Fig. 4 and 6. To test whether the association between poor SRH, poor mental-HRQoL and poor physical-HRQoL differed by sex in relation to LE8 and LS7 scores, a multiplicative interaction term was included in the models. Models were calculated considering 3 increasing levels of covariate adjustment: Model 1, unadjusted; Model 2, adjusted by sex, age, and study site (Linköping, Stockholm, Gothenburg, Lund/Malmö, Uppsala, Umeå); Model 3, adjusted by Model 2 covariates + educational status, marital status, and chronic disease (myocardial infarction, stroke, heart failure, peripheral arterial disease, chronic obstructive pulmonary disease, celiac disease, Crohn’s disease or ulcerative colitis disease, rheumatic disease, and cancer). Model 2 was considered as the main statistical model considering its clinical utility since it includes important covariates easy to collect in clinical practice while remaining relatively straightforward and easy to interpret.

Third, we further examined associations of LE8 and LS7 with poor mental-HRQoL and physical-HRQoL using Model 2 and similar reference points as described above (Fig. 3).

Fourth, we analyzed the capacity to discriminate poor SRH, poor mental-HRQoL and poor physical-HRQoL between LE8, LS7 (0–7) and LS7 (0–14) scores through receiver operating characteristic (ROC) curves and areas under the ROC curves (AUC) were compared with DeLong´s tests (Fig. 4).

Fifth, the associations of the different components of the LE8 score with SRH were studied through logistic regression models considering components as continuous variables (0–100) transformed to Z scores (Supplementary Fig. 3).

Finally, to examine the robustness of our main findings, we conducted a sensitivity analysis for the associations of LE8 and SRH by only including those with complete data on all LE8 components (instead of including those with data on ≥ 7 components).

Two-tailed p values < 0.05 were considered significant in all analyses and statistical analysis were performed using IBM SPSS Statistics Program (version 28.0, IBM Corp., NY, USA) and Stata 17 (StataCorp. 2021).