Interpretation of hypothesis generation results

Examine the results of the food frequency analysis using binomial probabilities below [original table available in the third tab of the Excel spreadsheet for Exercise 2 (Module 2 – Exercise 2)].

Note: In practice, binomial probabilities will be calculated for all food items, and then factors outlined in Parts 1 and 2 of this exercise will be taken into consideration.

The table shows that multiple food items are reported more commonly than would be expected: spinach, blueberries, almonds, walnuts, and sesame seeds.

 

Food Item

Confirmed Cases

  

Reference

Binomial Probability

Yes

Prob

No

DK

%Y+P

Foodbook Canada*

p-value

MEATS

Any chicken (not including deli meat)

3

0

3

1

50.0

85.6

0.0375

Any pork (not including deli meat)

1

2

3

1

50.0

55.1

0.3028

Any beef (not including deli meat)

1

1

4

1

33.3

78.4

0.0201

EGGS

Any eggs

2

3

2

0

71.4

80.7

0.2677

DAIRY PRODUCTS

Any dairy (excluding cheese)

3

1

3

0

57.1

84.6

0.0655

Non-dairy milk

3

0

3

1

50.0

No data

No data

Any cheese

4

0

3

0

57.1

88.8

0.00306

VEGETABLES

Tomatoes

3

1

3

0

57.1

72.9

0.1967

Any lettuce or leafy greens

4

1

1

1

83.3

82.4

0.4011

Iceberg

0

2

3

2

40.0

41.1

0.3452

Romaine

2

1

3

1

50.0

48.8

0.312

Spinach

4

0

1

2

80.0

28.4

0.0233

Sprouts

2

1

4

0

42.9

12.9

0.0432

Cucumbers

3

2

2

0

71.4

62.9

0.2846

Bell peppers

4

0

2

1

66.7

63.6

0.3252

Broccoli

3

0

3

1

50.0

55.5

0.3013

Cauliflower

4

0

3

0

57.1

33.0

0.1248

Mushrooms

4

0

3

0

57.1

50.0

0.2734

Zucchini

3

1

3

0

57.1

21.1

0.0341

FRUITS

Melons

3

0

3

1

50.0

39.7

0.2744

Apples

4

1

2

0

71.4

72.3

0.3183

Bananas

4

2

1

0

85.7

76.7

0.3321

Citrus fruits

4

0

3

0

57.1

65.0

0.2679

Any berries

5

0

2

0

71.4

65.2

0.2997

Strawberries

2

2

2

1

66.7

49.6

0.2306

Raspberries

2

0

3

2

40.0

27.5

0.2882

Blueberries

3

2

2

0

71.4

31.3

0.0298

Blackberries

3

1

3

0

57.1

10.5

0.003

Mangoes

4

0

3

0

57.1

15.7

0.0127

Pineapple

1

1

5

0

28.6

30.0

0.3177

NUTS & SEEDS

Peanuts

4

0

3

0

57.1

33.6

0.1306

Almonds

2

3

1

1

83.3

41.0

0.041

Walnuts

3

1

2

1

66.7

18.5

0.0117

Hazelnuts (filberts)

0

0

6

1

0.0

10.1

0.5279

Cashews

2

0

1

4

66.7

26.8

0.1577

Pecans

2

1

3

1

50.0

12.9

0.0284

Pistachios

0

0

4

3

0.0

No data

No data

Other nuts

1

0

3

3

25.0

No data

No data

Peanut butter

4

0

3

0

57.1

55.0

0.2918

Other nut butters/pastes/spreads

2

1

3

1

50.0

18.3

0.0668

Sunflower seeds

2

1

3

1

50.0

18.3

0.0668

Sesame seeds

2

2

2

1

66.7

17.1

0.0088

Chia seeds

3

2

2

0

71.4

No data

No data

Flax seeds

2

2

2

1

66.7

No data

No data

Other seeds

1

0

3

3

25.0

No data

No data

OTHER

Cold cereals

2

0

4

1

33.3

54.3

0.1929

Hot cereals

2

0

2

3

50.0

28.5

0.2491

Vegetarian/Vegan

2

0

3

2

40.0

No data

No data

Supplements

3

0

4

0

42.9

28.2

0.2086

*Based on the Foodbook Survey, 2015, Public Health Agency of Canada

 Question 2-9: What do these results mean? Why are multiple items identified?

Items that are reported more commonly than expected should be examined in further detail and assessed as a possible source of the outbreak (e.g., for packaged goods such as frozen berries, did cases report the same brand? For general produce items such as tomatoes, did cases report a particular type, e.g., cherry?). Statistically, with so many questions on a hypothesis-generating questionnaire, there will be some items that come up by chance alone (especially with a small sample size like the one in this case study).

Additionally, the outbreak under investigation represents a unique case demographic – there is one vegetarian and one vegan, as well as cases reporting diets rich in fresh produce. The produce items and nuts and seeds may be flagging because these are items typically consumed by this case demographic. Exposures may also flag if they are reported very infrequently compared to expected – it is important to look at the total number of cases reporting an exposure, while always keeping in mind what we discussed earlier around foods that might be more difficult to recall.  

On the other hand, it is important to keep in mind that some foods with high expected consumption levels (e.g., any eggs) may not flag statistically, but could still be potential sources. It is important to look for commonalities among commonly reported exposures.

Although expected consumption data were not available for chia and flax seeds, a high proportion of cases reported consuming these food items, suggesting these food items may be of interest as potential sources of the outbreak.

Question 2-10: If data from the Foodbook Study, or a similar study, were not available, what other studies could be conducted to help identify foods of interest? Why is the Foodbook study preferable in this situation?

An analytic study, such as a case-control study, could be conducted in place of the Foodbook study in order to compare the food consumption of cases to the food consumption of the general population (controls).

Use of an analytic study such as a case control study is not commonly used in a national outbreak investigation. These studies are costly to run, and take time. While analytic studies are very valuable in other situations, in this case the Foodbook data is readily available and representative of the Canadian population.

Further reading on analytic studies is available here

<Previous Next>