Longitudinal clustering of health behaviours and their association with multimorbidity: Evidence from Understanding Society (UKHLS)

Using longitudinal data from the UK Household Longitudinal Study, this study identifies seven stable clusters of health behaviors across adulthood and reveals a counterintuitive association where the "Overall Low-risk" behavioral cluster, characterized by older women with lower socioeconomic status, exhibits the highest prevalence of multimorbidity, highlighting the critical importance of considering sociodemographic context alongside behavioral patterns.

The Gist

Imagine your health isn't just a single light switch that is either "on" (healthy) or "off" (sick). Instead, think of it like a complex orchestra. You have four main instruments playing the tune of your daily life: Smoking, Nutrition, Alcohol, and Physical Activity (the researchers call this the "SNAP" band).

For a long time, doctors and scientists have listened to these instruments one by one. They'd say, "Oh, the smoking instrument is too loud," or "The exercise drum is too quiet." But this new study suggests that to really understand the music of your health, we need to listen to how all four instruments play together over many years.

Here is the story of what the researchers found, using some everyday metaphors:

1. The "Group Hug" vs. The "Solo Act"

The researchers looked at data from over 18,000 British adults (ages 16 and up) over seven years. Instead of looking at individuals, they grouped people into "tribes" based on how their four health instruments played together.

They found seven distinct tribes:

• The Low-Risk Squad: People who mostly eat well, don't smoke, and move around.
• The Couch Potatoes: People who eat okay but don't move enough.
• The Heavy Drinkers: People who enjoy a few too many drinks.
• The "Bad Habits" Combo: People who smoke, drink, and don't eat well.
• And a few others in between.

Think of these tribes like different playlists. Once you are on a playlist, you tend to stay there. If you were on the "Heavy Drinking" playlist in 2015, you were likely still on it in 2020. These habits are sticky!

2. The Great Plot Twist (The "Healthy" Paradox)

Here is where the story gets weird and counterintuitive.

Usually, we assume that if you are on the "Low-Risk Playlist" (no smoking, good food, exercise), you will be the healthiest person in the room. You'd expect them to have the fewest diseases.

But the data told a different story.

• The "Low-Risk" Tribe: Surprisingly, this group had the highest number of people with multiple chronic diseases (multimorbidity).
• The "Hazardous" Tribe: The group that drank too much, ate poorly, and didn't exercise actually had fewer diseases than you might expect.

Why? The "Who's in the Room" Factor.
Imagine two different parties:

• Party A (The Low-Risk Group): This party is mostly filled with older women who have had a hard life, lower incomes, and less education. They are trying their best to be healthy, but they are carrying the heavy baggage of age and poverty. Their "low-risk" habits are fighting a losing battle against their life circumstances.
• Party B (The Hazardous Group): This party is mostly filled with wealthy, highly educated people. Even though they are drinking too much and eating junk, their "safety net" of money, good healthcare, and stress-free jobs is protecting them from getting sick right now.

It's like two cars driving down a road.

• Car A is a brand new, shiny sports car (great habits), but it's being driven by an elderly driver on a bumpy, pothole-filled road (poverty/age). It's still getting damaged.
• Car B is an old, rusty truck with a flat tire (bad habits), but it's being driven by a pro racer on a smooth, paved highway (wealth/education). It's holding up surprisingly well.

3. The Big Takeaway

The main lesson here is that you can't judge a book by its cover, or a person's health just by their habits.

If you see someone who smokes and eats junk food, don't automatically assume they are the sickest person in the room. If you see someone who runs marathons and eats kale, don't assume they are immune to illness.

The researchers are saying: Look at the whole picture. To understand why people get sick, you have to listen to the entire orchestra—the habits plus the background music of their income, education, and age.

In short: Your habits matter, but your life story matters even more. To fix health problems, we need to help people not just change their playlists, but also fix the road they are driving on.

Technical Summary

1. Problem Statement

Health behaviors—specifically Smoking, Nutrition, Alcohol consumption, and Physical activity (collectively known as SNAP)—are established major risk factors for multimorbidity (the co-occurrence of two or more chronic conditions). However, existing literature often analyzes these behaviors in isolation rather than examining their synergistic effects as combined patterns.

While a previous study by Suhag et al. identified specific SNAP behavior clusters among older adults (aged ≥50), it remains unclear whether these patterns generalize across the broader adult lifespan. Furthermore, the relationship between these longitudinal behavioral clusters and multimorbidity prevalence, when adjusted for sociodemographic context, requires further investigation using independent, large-scale panel data.

2. Methodology

The study employed a rigorous quantitative approach using longitudinal data from the UK Household Longitudinal Study (Understanding Society).

• Data Source & Sample: The analysis utilized Waves 7, 9, 11, and 13 of the UKHLS. The final analytic sample consisted of 18,008 adults (aged ≥16) who participated in all seven waves, ensuring longitudinal consistency.
• Cluster Identification: The researchers applied Repeated-Measures Latent Class Analysis (RMLCA) to identify distinct subgroups (clusters) of individuals based on their SNAP behavior patterns across the four time points. This method accounts for the stability and evolution of behaviors over time.
• Statistical Modeling:
  • Multinomial Logistic Regression: Used to examine the association between the identified behavioral clusters and sociodemographic characteristics (e.g., age, gender, education, income).
  • Binomial Logistic Regression: Used to assess the association between the clusters and disease outcomes, specifically six distinct disease groups and the presence of multimorbidity.

3. Key Contributions

• Conceptual Replication & Extension: The study successfully replicated the clustering approach of Suhag et al. but extended the scope from older adults (≥50) to the entire adult population (≥16), providing a more comprehensive view of behavioral patterns across the lifespan.
• Longitudinal Behavioral Profiling: By using RMLCA, the study moved beyond cross-sectional snapshots to identify stable behavioral trajectories, offering a more robust classification of risk profiles.
• Contextual Integration: The research explicitly links behavioral clusters with sociodemographic variables, highlighting that health behaviors cannot be interpreted in a vacuum but are deeply embedded in social and economic contexts.

4. Key Results

The analysis identified seven distinct longitudinal clusters of SNAP behaviors, which remained largely stable over time:

1. Overall Low-risk (20%): Low engagement in all negative behaviors.
2. Insufficiently active (18%): Primarily characterized by low physical activity.
3. Poor diet and Insufficiently active (23%): The largest cluster, combining dietary issues with inactivity.
4. Hazardous and Harmful drinkers (11%): Focused on alcohol misuse.
5. Hazardous drinkers, Insufficiently active and Poor diet (14%): A high-risk profile combining alcohol, diet, and activity issues.
6. Smokers and Drinkers (5%): Combined tobacco and alcohol use.
7. Smokers (9%): Primarily tobacco use.

Counterintuitive Findings on Multimorbidity:
The study revealed a paradoxical relationship between behavioral risk profiles and disease prevalence:

• The "Overall Low-risk" cluster exhibited the highest prevalence of multimorbidity.
• Conversely, the "Hazardous drinkers, Insufficiently active and Poor diet" cluster showed lower prevalence across most conditions.

Sociodemographic Drivers:
This paradox was explained by the sociodemographic composition of the clusters:

• The "Overall Low-risk" group was predominantly composed of older women with lower education and income levels. Their high multimorbidity rates are likely driven by age and socioeconomic deprivation rather than the absence of SNAP risks.
• The "Hazardous drinkers, Insufficiently active and Poor diet" group was more likely to comprise individuals from the highest education and income groups. Despite engaging in multiple risky behaviors, their higher socioeconomic status may provide protective factors (e.g., better access to healthcare, nutrition, or stress management) that mitigate immediate multimorbidity risks.

5. Significance

• Methodological Insight: The study demonstrates the utility of cluster-analytic techniques in identifying population subgroups with unique behavioral and disease profiles, suggesting that traditional single-behavior risk assessments may be insufficient.
• Public Health Implications: The findings underscore the critical need to consider sociodemographic context when interpreting health behaviors. A "low-risk" behavioral profile does not guarantee low disease risk if the individual belongs to a vulnerable socioeconomic group.
• Policy Direction: Interventions targeting multimorbidity must be tailored not just to specific behaviors (e.g., smoking cessation) but to the specific intersection of behavior, age, and socioeconomic status. Policies addressing the "low-risk" but high-morbidity group (older, lower SES) may require different strategies than those targeting high-risk behavior groups with high SES.